metadata

license: bsd-2-clause
language:
  - fr
task_categories:
  - text-classification
task_ids:
  - multi-input-text-classification
size_categories:
  - 1K<n<10K

Dataset Card for Dataset Name

Dataset Description

Homepage:
Repository: https://github.com/mskandalis/daccord-dataset-contradictions
Paper: https://aclanthology.org/2023.jeptalnrecital-long.22/
Leaderboard:
Point of Contact:

Dataset Summary

The DACCORD dataset is an entirely new collection of 1034 sentence pairs annotated as a binary classification task for automatic detection of contradictions between sentences in French. Each pair of sentences receives a label according to whether or not the two sentences contradict each other. DACCORD currently covers the themes of Russia’s invasion of Ukraine in 2022, the Covid-19 pandemic, and the climate crisis. The sentences of the dataset were extracted from (or based on sentences from) AFP Factuel articles.

Supported Tasks and Leaderboards

The task of automatic detection of contradictions between sentences is a sentence-pair binary classification task. It can be viewed as a task related to both natural language inference task and misinformation detection task.

Dataset Structure

Data Fields

id: Index number.
premise: The translated premise in the target language.
hypothesis: The translated premise in the target language.
label: The classification label, with possible values 0 (compatibles), 1 (contradiction).
label_text: The classification label, with possible values compatibles (0), contradiction (1).
genre: a string feature .

Data Splits

theme	contradiction	compatible
Russian invasion	215	257
Covid-19	251	199
Climate change	49	63

Additional Information

Citation Information

BibTeX:

@inproceedings{skandalis-etal-2023-daccord,
    title = "{DACCORD} : un jeu de donn{\'e}es pour la D{\'e}tection Automatique d{'}{\'e}non{C}{\'e}s {CO}nt{R}a{D}ictoires en fran{\c{c}}ais",
    author = "Skandalis, Maximos  and
      Moot, Richard  and
      Robillard, Simon",
    booktitle = "Actes de CORIA-TALN 2023. Actes de la 30e Conf{\'e}rence sur le Traitement Automatique des Langues Naturelles (TALN), volume 1 : travaux de recherche originaux -- articles longs",
    month = "6",
    year = "2023",
    address = "Paris, France",
    publisher = "ATALA",
    url = "https://aclanthology.org/2023.jeptalnrecital-long.22",
    pages = "285--297",
    abstract = "La t{\^a}che de d{\'e}tection automatique de contradictions logiques entre {\'e}nonc{\'e}s en TALN est une t{\^a}che de classification binaire, o{\`u} chaque paire de phrases re{\c{c}}oit une {\'e}tiquette selon que les deux phrases se contredisent ou non. Elle peut {\^e}tre utilis{\'e}e afin de lutter contre la d{\'e}sinformation. Dans cet article, nous pr{\'e}sentons DACCORD, un jeu de donn{\'e}es d{\'e}di{\'e} {\`a} la t{\^a}che de d{\'e}tection automatique de contradictions entre phrases en fran{\c{c}}ais. Le jeu de donn{\'e}es {\'e}labor{\'e} est actuellement compos{\'e} de 1034 paires de phrases. Il couvre les th{\'e}matiques de l{'}invasion de la Russie en Ukraine en 2022, de la pand{\'e}mie de Covid-19 et de la crise climatique. Pour mettre en avant les possibilit{\'e}s de notre jeu de donn{\'e}es, nous {\'e}valuons les performances de certains mod{\`e}les de transformeurs sur lui. Nous constatons qu{'}il constitue pour eux un d{\'e}fi plus {\'e}lev{\'e} que les jeux de donn{\'e}es existants pour le fran{\c{c}}ais, qui sont d{\'e}j{\`a} peu nombreux. In NLP, the automatic detection of logical contradictions between statements is a binary classification task, in which a pair of sentences receives a label according to whether or not the two sentences contradict each other. This task has many potential applications, including combating disinformation. In this article, we present DACCORD, a new dataset dedicated to the task of automatically detecting contradictions between sentences in French. The dataset is currently composed of 1034 sentence pairs. It covers the themes of Russia{'}s invasion of Ukraine in 2022, the Covid-19 pandemic, and the climate crisis. To highlight the possibilities of our dataset, we evaluate the performance of some recent Transformer models on it. We conclude that our dataset is considerably more challenging than the few existing datasets for French.",
    language = "French",
}

ACL:

Maximos Skandalis, Richard Moot, and Simon Robillard. 2023. DACCORD : un jeu de données pour la Détection Automatique d’énonCés COntRaDictoires en français. In Actes de CORIA-TALN 2023. Actes de la 30e Conférence sur le Traitement Automatique des Langues Naturelles (TALN), volume 1 : travaux de recherche originaux -- articles longs, pages 285–297, Paris, France. ATALA.

Acknowledgements

This work was supported by the Defence Innovation Agency (AID) of the Directorate General of Armament (DGA) of the French Ministry of Armed Forces, and by the ICO, Institut Cybersécurité Occitanie, funded by Région Occitanie, France.