TFDS obsługuje teraz format Croissant 🥐 ! Przeczytaj dokumentację , aby dowiedzieć się więcej.

Ta strona została przetłumaczona przez Cloud Translation API.

tapako

Bibliografia:

wszystkie języki

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/all_languages')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1926192

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

af

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/af')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	307

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ar

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	6446

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

az

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/az')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	624

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

Być

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/be')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1512

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ber

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ber')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	67484

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

bg

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/bg')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	6324

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

miliard

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/bn')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1440

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

br

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/br')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2536

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ok

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ca')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	518

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

cbk

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/cbk')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	262

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

cmn

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	12549

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

cs

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/cs')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	6659

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

da

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/da')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	11220

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

de

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	125091

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

el

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	10072

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

pl

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	158053

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

eo

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/eo')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	207105

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

es

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	85064

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

i.t

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/et')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	241

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ue

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/eu')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	573

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

fi

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/fi')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	31753

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ks

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	116733

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

gl

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/gl')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	351

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

idź S

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/gos')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	279

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

On

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/he')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	68350

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

Cześć

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1913

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

godz

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/hr')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	505

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

hu

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/hu')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	67964

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

hy

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/hy')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	603

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

m.in

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ia')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2548

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ID

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/id')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1602

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tj

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ie')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	488

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

io

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/io')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	480

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

Jest

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/is')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1641

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

To

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/it')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	198919

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ja

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ja')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	44267

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

jbo

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/jbo')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2704

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

kab

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/kab')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	15944

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ko

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ko')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	503

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

kw

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/kw')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1328

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

la

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/la')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	6889

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

jeśli

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/lfn')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2313

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

Por

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/lt')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	8042

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

mk

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/mk')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	14678

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

Pan

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/mr')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	16413

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

uwaga

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/nb')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1094

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

nds

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/nds')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2633

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

nl

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/nl')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	23561

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

orw

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/orv')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	471

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tak

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ota')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	486

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

pes

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	4285

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

pl

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/pl')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	22391

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

pkt

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/pt')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	78430

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

rn

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/rn')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	648

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ro

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ro')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2092

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ru

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	251263

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

śl

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/sl')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	706

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

senior

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/sr')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	8175

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

sw

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/sv')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	7005

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tk

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/tk')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1165

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tl

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/tl')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1017

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tlh

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/tlh')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2804

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

toki

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/toki')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	3738

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tr

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	142088

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tt

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/tt')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	2398

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ug

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ug')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	1183

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

Wielka Brytania

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/uk')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	54431

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ur

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/ur')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	252

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

VI

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

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

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	962

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

vo

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/vo')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	328

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

wojna

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/war')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	327

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

wuu

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/wuu')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	408

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tak

Użyj następującego polecenia, aby załadować ten zestaw danych do TFDS:

ds = tfds.load('huggingface:tapaco/yue')

Opis :

A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a
crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular
linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links
between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent
filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows
that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or
near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000
sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists.

Licencja : Creative Commons Uznanie autorstwa 2.0 Generic
Wersja : 1.0.0
Podziały :

Podział	Przykłady
`'train'`	561

Cechy :

{
    "paraphrase_set_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "paraphrase": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "lists": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "tags": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "language": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}