lincoln/newsquadfr
Dataset Card for newsquadfr Dataset Summary newsquadfr is a small dataset created for Question Answering task. Contexts are paragraphs of articles extracted from nine online french newspaper during year 2020/2021. newsquadfr stands for Newspaper question answering dataset in french. inspired by Piaf and Squad dataset. 2 520 triplets context - question - answer. from datasets import load_dataset ds_name = 'lincoln/newsquadfr' # exemple 1 ds_newsquad =… See the full description on the dataset page: https://huggingface.co/datasets/lincoln/newsquadfr.
Dataset Card for newsquadfr
Table of Contents
- Dataset Description
- Dataset Summary
- Supported Tasks
- Languages
- Dataset Structure
- Data Instances
- Data Fields
- Data Splits
- Dataset Creation
- Curation Rationale
- Source Data
- Annotations
- 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
- Citation Information
Dataset Description
- Homepage: lincoln.fr
- Repository: github/Lincoln-France
- Paper: [Needs More Information]
- Leaderboard: [Needs More Information]
- Point of Contact: email
Dataset Summary
newsquadfr is a small dataset created for Question Answering task. Contexts are paragraphs of articles extracted from nine online french newspaper during year 2020/2021. newsquadfr stands for Newspaper question answering dataset in french. inspired by Piaf and Squad dataset. 2 520 triplets context - question - answer.
from datasets import load_dataset
ds_name = 'lincoln/newsquadfr'
# exemple 1
ds_newsquad = load_dataset(ds_name)
# exemple 2
data_files = {'train': 'train.json', 'test': 'test.json', 'valid': 'valid.json'}
ds_newsquad = load_dataset(ds_name, data_files=data_files)
# exemple 3
ds_newsquad = load_dataset(ds_name, data_files=data_files, split="valid+test")(train set)
Supported Tasks and Leaderboards
- extractive-qa
- open-domain-qa
Languages
Fr-fr
Dataset Structure
Data Instances
{'answers': {'answer_start': [53], 'text': ['manSuvre "agressive']},
'article_id': 34138,
'article_title': 'Caricatures, Libye, Haut-Karabakh... Les six dossiers qui '
'opposent Emmanuel Macron et Recep Tayyip Erdogan.',
'article_url': 'https://www.francetvinfo.fr/monde/turquie/caricatures-libye-haut-karabakh-les-six-dossiers-qui-opposent-emmanuel-macron-et-recep-tayyip-erdogan_4155611.html#xtor=RSS-3-[france]',
'context': 'Dans ce contexte déjà tendu, la France a dénoncé une manSuvre '
'"agressive" de la part de frégates turques à l\'encontre de l\'un '
"de ses navires engagés dans une mission de l'Otan, le 10 juin. "
'Selon Paris, la frégate Le Courbet cherchait à identifier un '
'cargo suspecté de transporter des armes vers la Libye quand elle '
'a été illuminée à trois reprises par le radar de conduite de tir '
"de l'escorte turque.",
'id': '2261',
'paragraph_id': 201225,
'question': "Qu'est ce que la France reproche à la Turquie?",
'website': 'francetvinfo'}Data Fields
answers: a dictionary feature containing:text: astringfeature.answer_start: aint64feature.article_id: aint64feature.article_title: a string feature.article_url: a string feature.context: astringfeature.id: astringfeature.paragraph_id: aint64feature.question: astringfeature.website: astringfeature.
Data Splits
Dataset Creation
Curation Rationale
[Needs More Information]
Source Data
Initial Data Collection and Normalization
Paragraphs were chosen according to theses rules:
- parent article must have more than 71% ASCII characters
- paragraphs size must be between 170 and 670 characters
- paragraphs shouldn't contain "A LIRE" or "A VOIR AUSSI"
Then, we stratified our original dataset to create this dataset according to :
- website
- number of named entities
- paragraph size
Who are the source language producers?
[Needs More Information]
Annotations
Annotation process
Using Piaf annotation tools. Three different persons mostly.
Who are the annotators?
Lincoln
Personal and Sensitive Information
[Needs More Information]
Considerations for Using the Data
Social Impact of Dataset
[Needs More Information]
Discussion of Biases
- Annotation is not well controlled
- asking question on news is biaised
Other Known Limitations
[Needs More Information]
Additional Information
Dataset Curators
[Needs More Information]
Licensing Information
https://creativecommons.org/licenses/by-nc-sa/4.0/deed.fr
Citation Information
[Needs More Information]
