Riferimenti:
tutte_lingue
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/all_languages')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1926192 |
- Caratteristiche :
{
"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"
}
}
ef
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/af')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 307 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ar')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 6446 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/az')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 624 |
- Caratteristiche :
{
"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"
}
}
Essere
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/be')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1512 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ber')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 67484 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/bg')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 6324 |
- Caratteristiche :
{
"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"
}
}
miliardo
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/bn')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1440 |
- Caratteristiche :
{
"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"
}
}
fratello
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/br')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2536 |
- Caratteristiche :
{
"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"
}
}
ca
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ca')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 518 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/cbk')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 262 |
- Caratteristiche :
{
"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"
}
}
cmq
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/cmn')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 12549 |
- Caratteristiche :
{
"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"
}
}
c.s
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/cs')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 6659 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/da')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 11220 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/de')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 125091 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/el')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 10072 |
- Caratteristiche :
{
"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"
}
}
en
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/en')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 158053 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/eo')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 207105 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/es')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 85064 |
- Caratteristiche :
{
"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"
}
}
et
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/et')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 241 |
- Caratteristiche :
{
"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"
}
}
Unione Europea
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/eu')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 573 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/fi')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 31753 |
- Caratteristiche :
{
"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"
}
}
fr
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/fr')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 116733 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/gl')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 351 |
- Caratteristiche :
{
"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"
}
}
va
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/gos')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 279 |
- Caratteristiche :
{
"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"
}
}
Lui
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/he')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 68350 |
- Caratteristiche :
{
"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"
}
}
CIAO
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/hi')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1913 |
- Caratteristiche :
{
"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"
}
}
ora
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/hr')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 505 |
- Caratteristiche :
{
"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"
}
}
eh
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/hu')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 67964 |
- Caratteristiche :
{
"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"
}
}
ciao
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/hy')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 603 |
- Caratteristiche :
{
"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"
}
}
e.a
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ia')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2548 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/id')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1602 |
- Caratteristiche :
{
"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"
}
}
cioè
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ie')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 488 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/io')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 480 |
- Caratteristiche :
{
"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"
}
}
È
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/is')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1641 |
- Caratteristiche :
{
"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"
}
}
Esso
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/it')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 198919 |
- Caratteristiche :
{
"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"
}
}
sì
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ja')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 44267 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/jbo')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2704 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/kab')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 15944 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ko')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 503 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/kw')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1328 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/la')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 6889 |
- Caratteristiche :
{
"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"
}
}
lfn
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/lfn')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2313 |
- Caratteristiche :
{
"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"
}
}
lt
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/lt')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 8042 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/mk')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 14678 |
- Caratteristiche :
{
"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"
}
}
Sig
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/mr')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 16413 |
- Caratteristiche :
{
"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"
}
}
n.b
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/nb')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1094 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/nds')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2633 |
- Caratteristiche :
{
"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"
}
}
n.l
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/nl')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 23561 |
- Caratteristiche :
{
"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"
}
}
orv
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/orv')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 471 |
- Caratteristiche :
{
"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"
}
}
ota
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ota')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 486 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/pes')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 4285 |
- Caratteristiche :
{
"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"
}
}
per favore
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/pl')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 22391 |
- Caratteristiche :
{
"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"
}
}
punto
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/pt')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 78430 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/rn')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 648 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ro')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2092 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ru')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 251263 |
- Caratteristiche :
{
"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"
}
}
sl
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/sl')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 706 |
- Caratteristiche :
{
"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"
}
}
sr
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/sr')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 8175 |
- Caratteristiche :
{
"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"
}
}
sv
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/sv')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 7005 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/tk')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1165 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/tl')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1017 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/tlh')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2804 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/toki')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 3738 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/tr')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 142088 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/tt')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 2398 |
- Caratteristiche :
{
"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"
}
}
ehm
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ug')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 1183 |
- Caratteristiche :
{
"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"
}
}
Regno Unito
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/uk')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 54431 |
- Caratteristiche :
{
"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"
}
}
il tuo
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/ur')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 252 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/vi')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 962 |
- Caratteristiche :
{
"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"
}
}
voi
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/vo')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 328 |
- Caratteristiche :
{
"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"
}
}
guerra
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/war')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 327 |
- Caratteristiche :
{
"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
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/wuu')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 408 |
- Caratteristiche :
{
"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"
}
}
sì
Utilizzare il comando seguente per caricare questo set di dati in TFDS:
ds = tfds.load('huggingface:tapaco/yue')
- Descrizione :
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.
- Licenza : Creative Commons Attribuzione 2.0 Generica
- Versione : 1.0.0
- Divide :
Diviso | Esempi |
---|---|
'train' | 561 |
- Caratteristiche :
{
"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"
}
}