The paper presents ‘Ta keed’, an Arabic automatic fact-checking system that not only classifies information as true or false but also provides explanations for its classifications. The system utilises two main components: information retrieval and LLM-based claim verification. The system’s effectiveness was tested using the ArFactEx dataset, achieving an F1 score of 0.72. The paper also discusses the impact of the number of snippets retrieved on claim classification accuracy. The system’s generated explanations were compared with gold-standard explanations, recommending evaluation using semantic similarities.
Publication date: 26 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.14067
