Dataset Card for XNLI Parallel Corpus

Dataset Summary

Supported Tasks and Leaderboards

Binary mode classification (spoken vs written)

Languages

English
German
French

Dataset Structure

Data Instances

{ 'text': "And he said , Mama , I 'm home .", 'label': 0 }

Data Fields

text: sentence
label: binary label of text (0: spoken 1: written)

Data Splits

train: 830
test: 1669

Other Statistics

Vocabulary Size

English
- train: 4363
- test: 7128
German
- train: 5070
- test: 8601
French
- train: 4881
- test: 7935

Average Sentence Length

English
- train: 20.689156626506023
- test: 20.75254643499101
German
- train: 20.367469879518072
- test: 20.639904134212102
French
- train: 23.455421686746988
- test: 23.731575793888556

Label Split

train:
- 0: 166
- 1: 664
test:
- 0: 334
- 1: 1335

Out-of-vocabulary words in model

English
- BERT (bert-base-uncased)
  - train: 800
  - test: 1638
- mBERT (bert-base-multilingual-uncased)
  - train: 1347
  - test: 2693
- German BERT (bert-base-german-dbmdz-uncased)
  - train: 3228
  - test: 5581
- flauBERT (flaubert-base-uncased)
  - train: 4363
  - test: 7128
German
- BERT (bert-base-uncased)
  - train: 4285
  - test: 7387
- mBERT (bert-base-multilingual-uncased)
  - train: 3126
  - test: 5863
- German BERT (bert-base-german-dbmdz-uncased)
  - train: 2033
  - test: 3938
- flauBERT (flaubert-base-uncased)
  - train: 5069
  - test: 8600
French
- BERT (bert-base-uncased)
  - train: 3784
  - test: 6289
- mBERT (bert-base-multilingual-uncased)
  - train: 2847
  - test: 5084
- German BERT (bert-base-german-dbmdz-uncased)
  - train: 4212
  - test: 6964
- flauBERT (flaubert-base-uncased)
  - train: 4881
  - test: 7935

Dataset Creation

Curation Rationale

N/A

Source Data

https://github.com/facebookresearch/XNLI

Here is the citation for the original XNLI paper.

@InProceedings{conneau2018xnli,
  author = "Conneau, Alexis
        and Rinott, Ruty
        and Lample, Guillaume
        and Williams, Adina
        and Bowman, Samuel R.
        and Schwenk, Holger
        and Stoyanov, Veselin",
  title = "XNLI: Evaluating Cross-lingual Sentence Representations",
  booktitle = "Proceedings of the 2018 Conference on Empirical Methods
               in Natural Language Processing",
  year = "2018",
  publisher = "Association for Computational Linguistics",
  location = "Brussels, Belgium",
}

Initial Data Collection and Normalization

N/A

Who are the source language producers?

N/A

Annotations

Annotation process

N/A

Who are the annotators?

N/A

Personal and Sensitive Information

N/A

Considerations for Using the Data

Social Impact of Dataset

N/A

Discussion of Biases

N/A

Other Known Limitations

N/A

Additional Information

Dataset Curators

N/A

Licensing Information

N/A

Citation Information

Contributions

N/A