エクストリーム

参考文献:

• コード
• ハグフェイス

XNLI

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XNLI')

• 説明：

The Cross-lingual Natural Language Inference (XNLI) corpus is a crowd-sourced collection of 5,000 test and
2,500 dev pairs for the MultiNLI corpus. The pairs are annotated with textual entailment and translated into
14 languages: French, Spanish, German, Greek, Bulgarian, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese,
Hindi, Swahili and Urdu. This results in 112.5k annotated pairs. Each premise can be associated with the
corresponding hypothesis in the 15 languages, summing up to more than 1.5M combinations. The corpus is made to
evaluate how to perform inference in any language (including low-resources ones like Swahili or Urdu) when only
English NLI data is available at training time. One solution is cross-lingual sentence encoding, for which XNLI
is an evaluation benchmark.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gold_label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ティディカ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tydiqa')

• 説明：

Gold passage task (GoldP): Given a passage that is guaranteed to contain the
             answer, predict the single contiguous span of characters that answers the question. This is more similar to
             existing reading comprehension datasets (as opposed to the information-seeking task outlined above).
             This task is constructed with two goals in mind: (1) more directly comparing with prior work and (2) providing
             a simplified way for researchers to use TyDi QA by providing compatibility with existing code for SQuAD 1.1,
             XQuAD, and MLQA. Toward these goals, the gold passage task differs from the primary task in several ways:
             only the gold answer passage is provided rather than the entire Wikipedia article;
             unanswerable questions have been discarded, similar to MLQA and XQuAD;
             we evaluate with the SQuAD 1.1 metrics like XQuAD; and
            Thai and Japanese are removed since the lack of whitespace breaks some tools.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

分隊

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/SQuAD')

• 説明：

Stanford Question Answering Dataset (SQuAD) is a reading comprehension     dataset, consisting of questions posed by crowdworkers on a set of Wikipedia     articles, where the answer to every question is a segment of text, or span,     from the corresponding reading passage, or the question might be unanswerable.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.af

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.af')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ar')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.bg

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.bg')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.bn

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.bn')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.de')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.el

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.el')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.jp

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.en')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.es')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.et

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.et')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.eu

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.eu')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.fa

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.fa')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.fi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.fi')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.fr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.fr')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.he

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.he')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.hi')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.hu

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.hu')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.id

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.id')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.it

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.it')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ja

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ja')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.jv

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.jv')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ka

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ka')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.kk

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.kk')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ko

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ko')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ml

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ml')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.mr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.mr')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ms

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ms')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.my

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.my')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.nl

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.nl')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.pt

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.pt')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ru

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ru')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.sw

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.sw')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ta

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ta')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.te

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.te')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.th

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.th')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.tl

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.tl')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.tr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.tr')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.ur

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.ur')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.vi')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.yo

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.yo')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

PAN-X.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAN-X.zh')

• 説明：

The WikiANN dataset (Pan et al. 2017) is a dataset with NER annotations for PER, ORG and LOC. It has been
constructed using the linked entities in Wikipedia pages for 282 different languages including Danish. The dataset
can be loaded with the DaNLP package:
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "ner_tags": {
        "feature": {
            "num_classes": 7,
            "names": [
                "O",
                "B-PER",
                "I-PER",
                "B-ORG",
                "I-ORG",
                "B-LOC",
                "I-LOC"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "langs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.ja

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.ar.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.ar.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.en

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.de.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.de.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.en

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.vi.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.vi.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.en

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.zh.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.zh.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.en

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.en.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.en.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.en

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.es.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.es.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.hi.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.ar')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.hi.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.de')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.hi.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.vi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.hi.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.zh')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.hi.en

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.en')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.hi.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.es')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

MLQA.こんにちは。

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/MLQA.hi.hi')

• 説明：

MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance.
MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic,
German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between
4 different languages on average.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.ar

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.ar')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.de')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.vi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.vi')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.zh')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.jp

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.en')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.es')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.hi

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.hi')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.el

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.el')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.ru

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.ru')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.th

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.th')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

XQuAD.tr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/XQuAD.tr')

• 説明：

XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating cross-lingual question
answering performance. The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with their professional translations into
ten languages: Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi. Consequently,
the dataset is entirely parallel across 11 languages.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            },
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

bucc18.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/bucc18.de')

• 説明：

Building and Using Comparable Corpora

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

bucc18.fr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/bucc18.fr')

• 説明：

Building and Using Comparable Corpora

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

bucc18.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/bucc18.zh')

• 説明：

Building and Using Comparable Corpora

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

bucc18.ru

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/bucc18.ru')

• 説明：

Building and Using Comparable Corpora

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.de

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.de')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.jp

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.en')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.es

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.es')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.fr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.fr')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.ja

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.ja')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.ko

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.ko')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

PAWS-X.zh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/PAWS-X.zh')

• 説明：

This dataset contains 23,659 human translated PAWS evaluation pairs and 296,406 machine translated training
pairs in six typologically distinct languages: French, Spanish, German, Chinese, Japanese, and Korean. All
translated pairs are sourced from examples in PAWS-Wiki.
The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "sentence1": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence2": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "label": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.afr

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.afr')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.アラ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.ara')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ・ベン

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.ben')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.ブル

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.bul')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ・デウ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.deu')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.cmn

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.cmn')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバエル

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.ell')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.est

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.est')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ・エウス

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.eus')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバフィン

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.fin')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.フラ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.fra')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.ヘブ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.heb')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバヒン

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.hin')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ・フン

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.hun')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバインド

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.ind')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.イタ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.ita')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.jav

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.jav')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.jpn

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.jpn')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバカット

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.kat')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ・カズ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.kaz')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.コル

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.kor')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバマル

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.mal')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.マール

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.mar')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.nld

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.nld')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバペス

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.pes')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.por

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.por')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ・ラス

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.rus')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバスパ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.spa')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.swh

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.swh')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバタム

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.tam')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバテル

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.tel')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.tgl

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.tgl')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.tha')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.トゥル

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.tur')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.ウルド

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.urd')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

タトエバ.ヴィー

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/tatoeba.vie')

• 説明：

his data is extracted from the Tatoeba corpus, dated Saturday 2018/11/17.
For each languages, we have selected 1000 English sentences and their translations, if available. Please check
this paper for a description of the languages, their families and scripts as well as baseline results.
Please note that the English sentences are not identical for all language pairs. This means that the results are
not directly comparable across languages. In particular, the sentences tend to have less variety for several
low-resource languages, e.g. "Tom needed water", "Tom needs water", "Tom is getting water", ...

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "source_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_lang": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

udpos.アフリカーンス語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Afrikaans')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.アラビア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Arabic')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.バスク語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Basque')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ブルガリア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Bulgarian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.オランダ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Dutch')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.英語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.English')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.エストニア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Estonian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.フィンランド語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Finnish')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.フランス語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.French')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ドイツ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.German')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ギリシャ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Greek')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ヘブライ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Hebrew')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ヒンディー語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Hindi')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ハンガリー語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Hungarian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.インドネシア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Indonesian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.イタリア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Italian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.日本語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Japanese')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.カザフ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Kazakh')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.韓国語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Korean')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.中国語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Chinese')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.マラーティー語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Marathi')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ペルシア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Persian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ポルトガル語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Portuguese')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ロシア語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Russian')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.スペイン語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Spanish')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.タガログ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Tagalog')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.タミル語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Tamil')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.テルグ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Telugu')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.タイ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Thai')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.トルコ語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Turkish')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ウルドゥー語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Urdu')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ベトナム語

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Vietnamese')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

udpos.ヨルバ

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:xtreme/udpos.Yoruba')

• 説明：

Universal Dependencies (UD) is a framework for consistent annotation of grammar (parts of speech, morphological
features, and syntactic dependencies) across different human languages. UD is an open community effort with over 200
contributors producing more than 100 treebanks in over 70 languages. If you’re new to UD, you should start by reading
the first part of the Short Introduction and then browsing the annotation guidelines.

The Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark is a benchmark for the evaluation of
the cross-lingual generalization ability of pre-trained multilingual models. It covers 40 typologically diverse languages
(spanning 12 language families) and includes nine tasks that collectively require reasoning about different levels of
syntax and semantics. The languages in XTREME are selected to maximize language diversity, coverage in existing tasks,
and availability of training data. Among these are many under-studied languages, such as the Dravidian languages Tamil
(spoken in southern India, Sri Lanka, and Singapore), Telugu and Malayalam (spoken mainly in southern India), and the
Niger-Congo languages Swahili and Yoruba, spoken in Africa.

• ライセンス: 不明なライセンス
• バージョン: 1.0.0
• 分割:

• 特徴：

{
    "tokens": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "pos_tags": {
        "feature": {
            "num_classes": 17,
            "names": [
                "ADJ",
                "ADP",
                "ADV",
                "AUX",
                "CCONJ",
                "DET",
                "INTJ",
                "NOUN",
                "NUM",
                "PART",
                "PRON",
                "PROPN",
                "PUNCT",
                "SCONJ",
                "SYM",
                "VERB",
                "X"
            ],
            "names_file": null,
            "id": null,
            "_type": "ClassLabel"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}