Communication Dans Un Congrès Année : 2025

Leveraging Argumentation Schemes in Justification Generation for Automated Fact-checking

Résumé

Automatically generated justifications have become a standard component in automated fact-checking systems, serving as input to claim verification models. While significant progress has been made in retrieving relevant evidence for justification generation, the semantic of the generated justifications remain largely underexplored despite their critical role in enhancing verification effectiveness. In this paper, we introduce WACID, Walton-scheme-based Argumentative and Critical Interactive Dialogue, a novel dialogical framework that integrates Argumentation Schemes into the justification generation process for the first time, enabling the production of more finegrained argumentation-based justifications by leveraging schemespecific critical questions. We evaluate our approach by feeding the generated justifications into three recent claim verification models, achieving state-of-the-art performance across multiple benchmark datasets compared to three competitive baselines. Additionally, we show that our method can serve as a complementary knowledge layer, boosting the outputs of existing generation methods and yielding an average F1 score improvement of 0.07 when paired with the same verification models.

Fichier principal
Vignette du fichier
WI450.pdf (782.48 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05370762 , version 1 (18-11-2025)

Licence

Identifiants

  • HAL Id : hal-05370762 , version 1

Citer

Xiaoou Wang, Elena Cabrio, Serena Villata. Leveraging Argumentation Schemes in Justification Generation for Automated Fact-checking. The 24th IEEE/WIC International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT 2025), Nov 2025, Londres, United Kingdom. ⟨hal-05370762⟩
72 Consultations
159 Téléchargements

Partager

  • More