gem

References:

mlsum_de

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/mlsum_de')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_covid' 5058
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 10695
'train' 220748
'validation' 11392
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "text": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "topic": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "url": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "date": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

mlsum_es

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/mlsum_es')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_covid' 1938
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 13366
'train' 259888
'validation' 9977
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "text": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "topic": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "url": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "date": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_es_en_v0

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_es_en_v0')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 19797
'train' 79515
'validation' 8835
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_ru_en_v0

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_ru_en_v0')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 9094
'train' 36898
'validation' 4100
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_tr_en_v0

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_tr_en_v0')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 808
'train' 3193
'validation' 355
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_vi_en_v0

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_vi_en_v0')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 2167
'train' 9206
'validation' 1023
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_arabic_ar

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_arabic_ar')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 5841
'train' 20441
'validation' 2919
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "ar",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "ar",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_chinese_zh

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_chinese_zh')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 3775
'train' 13211
'validation' 1886
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "zh",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "zh",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_czech_cs

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_czech_cs')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 1438
'train' 5033
'validation' 718
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "cs",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "cs",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_dutch_nl

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_dutch_nl')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 6248
'train' 21866
'validation' 3123
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "nl",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "nl",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_english_en

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_english_en')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 28614
'train' 99020
'validation' 13823
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "en",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "en",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_french_fr

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_french_fr')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 12731
'train' 44556
'validation' 6364
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "fr",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "fr",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_german_de

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_german_de')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 11669
'train' 40839
'validation' 5833
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "de",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "de",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_hindi_hi

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_hindi_hi')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 1984
'train' 6942
'validation' 991
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "hi",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "hi",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_indonesian_id

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_indonesian_id')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 9497
'train' 33237
'validation' 4747
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "id",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "id",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_italian_it

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_italian_it')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 10189
'train' 35661
'validation' 5093
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "it",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "it",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_japanese_ja

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_japanese_ja')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 2530
'train' 8853
'validation' 1264
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "ja",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "ja",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_korean_ko

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_korean_ko')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 2436
'train' 8524
'validation' 1216
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "ko",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "ko",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_portuguese_pt

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_portuguese_pt')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 16331
'train' 57159
'validation' 8165
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "pt",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "pt",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_russian_ru

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_russian_ru')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 10580
'train' 37028
'validation' 5288
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "ru",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "ru",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_spanish_es

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_spanish_es')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 22632
'train' 79212
'validation' 11316
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "es",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "es",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_thai_th

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_thai_th')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 2950
'train' 10325
'validation' 1475
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "th",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "th",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_turkish_tr

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_turkish_tr')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 900
'train' 3148
'validation' 449
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "tr",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "tr",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

wiki_lingua_vietnamese_vi

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_lingua_vietnamese_vi')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 3917
'train' 13707
'validation' 1957
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source_aligned": {
        "languages": [
            "vi",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "target_aligned": {
        "languages": [
            "vi",
            "en"
        ],
        "id": null,
        "_type": "Translation"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

xsum

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/xsum')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_backtranslation' 500
'challenge_test_bfp_02' 500
'challenge_test_bfp_05' 500
'challenge_test_covid' 401
'challenge_test_nopunc' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 1166
'train' 23206
'validation' 1117
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "xsum_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "document": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

common_gen

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/common_gen')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 1497
'train' 67389
'validation' 993
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "concept_set_id": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "concepts": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

cs_restaurants

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/cs_restaurants')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 842
'train' 3569
'validation' 781
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "dialog_act": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "dialog_act_delexicalized": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target_delexicalized": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

dart

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/dart')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'test' 5097
'train' 62659
'validation' 2768
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "dart_id": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "tripleset": [
        [
            {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        ]
    ],
    "subtree_was_extended": {
        "dtype": "bool",
        "id": null,
        "_type": "Value"
    },
    "target_sources": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

e2e_nlg

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/e2e_nlg')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 4693
'train' 33525
'validation' 4299
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "meaning_representation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

totto

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/totto')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 7700
'train' 121153
'validation' 7700
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "totto_id": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "table_page_title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "table_webpage_url": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "table_section_title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "table_section_text": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "table": [
        [
            {
                "column_span": {
                    "dtype": "int32",
                    "id": null,
                    "_type": "Value"
                },
                "is_header": {
                    "dtype": "bool",
                    "id": null,
                    "_type": "Value"
                },
                "row_span": {
                    "dtype": "int32",
                    "id": null,
                    "_type": "Value"
                },
                "value": {
                    "dtype": "string",
                    "id": null,
                    "_type": "Value"
                }
            }
        ]
    ],
    "highlighted_cells": [
        [
            {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        ]
    ],
    "example_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_annotations": [
        {
            "original_sentence": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "sentence_after_deletion": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "sentence_after_ambiguity": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "final_sentence": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            }
        }
    ],
    "overlap_subset": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

web_nlg_en

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/web_nlg_en')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_numbers' 500
'challenge_test_scramble' 500
'challenge_train_sample' 502
'challenge_validation_sample' 499
'test' 1779
'train' 35426
'validation' 1667
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "input": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "category": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "webnlg_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

web_nlg_ru

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/web_nlg_ru')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 501
'challenge_validation_sample' 500
'test' 1102
'train' 14630
'validation' 790
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "input": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "category": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "webnlg_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

wiki_auto_asset_turk

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/wiki_auto_asset_turk')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_asset_backtranslation' 359
'challenge_test_asset_bfp02' 359
'challenge_test_asset_bfp05' 359
'challenge_test_asset_nopunc' 359
'challenge_test_turk_backtranslation' 359
'challenge_test_turk_bfp02' 359
'challenge_test_turk_bfp05' 359
'challenge_test_turk_nopunc' 359
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test_asset' 359
'test_turk' 359
'train' 483801
'validation' 20000
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "source": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}

schema_guided_dialog

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:gem/schema_guided_dialog')
  • Description:
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.

GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.

It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
  • License: CC-BY-SA-4.0
  • Version: 1.1.0
  • Splits:
Split Examples
'challenge_test_backtranslation' 500
'challenge_test_bfp02' 500
'challenge_test_bfp05' 500
'challenge_test_nopunc' 500
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 10000
'train' 164982
'validation' 10000
  • Features:
{
    "gem_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "gem_parent_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "dialog_acts": [
        {
            "act": {
                "num_classes": 18,
                "names": [
                    "AFFIRM",
                    "AFFIRM_INTENT",
                    "CONFIRM",
                    "GOODBYE",
                    "INFORM",
                    "INFORM_COUNT",
                    "INFORM_INTENT",
                    "NEGATE",
                    "NEGATE_INTENT",
                    "NOTIFY_FAILURE",
                    "NOTIFY_SUCCESS",
                    "OFFER",
                    "OFFER_INTENT",
                    "REQUEST",
                    "REQUEST_ALTS",
                    "REQ_MORE",
                    "SELECT",
                    "THANK_YOU"
                ],
                "id": null,
                "_type": "ClassLabel"
            },
            "slot": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "values": [
                {
                    "dtype": "string",
                    "id": null,
                    "_type": "Value"
                }
            ]
        }
    ],
    "context": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ],
    "dialog_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "service": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "turn_id": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "prompt": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "target": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "references": [
        {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        }
    ]
}