Detecting Fake News: Exploring Key Features in Multilingual Arabic Dialect Corpus - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2024

Detecting Fake News: Exploring Key Features in Multilingual Arabic Dialect Corpus

Résumé

As misinformation continues to spread rapidly on social media platforms identifying and stopping the dissemination of fake news has become an urgent need. In this article, we propose a deep learning approach leveraging keywords for feature extraction and classification of Arabic dialect fake news. Our method achieves an accuracy of 82.3% on a corpus comprising 3000 news articles in Algerian and Tunisian dialects, Modern Standard Arabic (MSA), French, and English, featuring instances of code-switching between these languages; as well as an accuracy of 93.7% on an English fake news corpus. Our experimentation shows that the shortcut learning problem that can arise when using keyword based features can be solved using regularization techniques. Our findings also show that our approach will achieve better performance on larger Arabic dialect corpora.
Fichier principal
Vignette du fichier
icalp_2023.pdf (1.15 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04578312 , version 1 (11-07-2024)

Identifiants

  • HAL Id : hal-04578312 , version 1

Citer

Hocini Abdelouahab, Kamel Smaïli. Detecting Fake News: Exploring Key Features in Multilingual Arabic Dialect Corpus. The 8th International Conference on Arabic Language Processing, Apr 2024, RABAT, Morocco. ⟨hal-04578312⟩
0 Consultations
1 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More