Datasets:
annotations_creators:
- no-annotation
language_creators:
- found
languages:
- fr-FR
licenses:
- cc-by-sa-4.0
multilinguality:
- monolingual
pretty_name: Wikitext-fr
size_categories:
- unknown
source_datasets:
- original
task_categories:
- sequence-modeling
task_ids:
- language-modeling
Dataset Card Creation Guide
Table of Contents
- Dataset Card Creation Guide
Dataset Description
- Repository: https://github.com/AntoineSimoulin/gpt-fr
- Paper: https://aclanthology.org/2021.jeptalnrecital-taln.24.pdf
Dataset Summary
Wikitext-fr language modeling dataset consists of over 70 million tokens extracted from the set of french Wikipedia articles that are classified as "quality articles" or "good articles". It is designed to mirror the english benchmark from Stephen Merity, Caiming Xiong, James Bradbury, and Richard Socher. 2016. Pointer Sentinel Mixture Models The dataset is available under the Creative Commons Attribution-ShareAlike License
Supported Tasks and Leaderboards
language-modeling: The dataset can be used to evaluate the generation abilites of a model. Success on this task is typically measured by achieving a low perplexity. The (model name currently achieves 12.9.
Languages
The dataset is in French.
Dataset Structure
Data Instances
The dataset consists in the agregation of paragraphs from wikipedia articles.
{
'paragraph': ...,
...
}
Data Fields
paragraph: This is a paragraph from the original wikipedia article.
Data Splits
The dataset is splited into a train/valid/test split.
| Tain (35) | Train (72) | Valid | Test | |
|---|---|---|---|---|
| Number of Documents | 2 126 | 5 902 | 60 | 60 |
| Number of tokens | 351 66 | 72 961 | 896 | 897 |
| Vocabulary size | 137 589 | 205 403 | ||
| Out of Vocabulary | 0.8% | 1.2% |
Dataset Creation
Curation Rationale
The dataset is created to evaluate French models with similart criteria than English.s
Source Data
Wikitext-fr language modeling dataset consists of over 70 million tokens extracted from the set of french Wikipedia articles that are classified as "quality articles" or "good articles". We did not apply specific pre-treatments as transformers models might use a dedicated tokenization.s
Initial Data Collection and Normalization
We used the Wikipedia API to collect the articles since cleaning Wikipedia articles from dumps is not a trivial task.
Personal and Sensitive Information
Considerations for Using the Data
Social Impact of Dataset
Discussion of Biases
Other Known Limitations
Additional Information
Dataset Curators
Licensing Information
The dataset is available under the Creative Commons Attribution-ShareAlike License
Citation Information
@inproceedings{simoulin:hal-03265900,
TITLE = {{Un mod{\`e}le Transformer G{\'e}n{\'e}ratif Pr{\'e}-entrain{\'e} pour le \_\_\_\_\_\_ fran{\c c}ais}},
AUTHOR = {Simoulin, Antoine and Crabb{\'e}, Benoit},
URL = {https://hal.archives-ouvertes.fr/hal-03265900},
BOOKTITLE = {{Traitement Automatique des Langues Naturelles}},
ADDRESS = {Lille, France},
EDITOR = {Denis, Pascal and Grabar, Natalia and Fraisse, Amel and Cardon, R{\'e}mi and Jacquemin, Bernard and Kergosien, Eric and Balvet, Antonio},
PUBLISHER = {{ATALA}},
PAGES = {246-255},
YEAR = {2021},
KEYWORDS = {fran{\c c}ais. ; GPT ; G{\'e}n{\'e}ratif ; Transformer ; Pr{\'e}-entra{\^i}n{\'e}},
PDF = {https://hal.archives-ouvertes.fr/hal-03265900/file/7.pdf},
HAL_ID = {hal-03265900},
HAL_VERSION = {v1},
}
Contributions
Thanks to @AntoineSimoulin for adding this dataset.