This paper presents the approach for the ‘Arabic AI Tasks Evaluation (ArAiEval) Shared Task 2023’, focusing on persuasion technique detection and disinformation detection. The authors experiment with several transformer-based models pre-trained on the Arabic language, fine-tuning these models on the provided dataset. They also employ ensembling to enhance the performance of the systems. The paper highlights the increasingly important role of machine learning in detecting disinformation and persuasion techniques in today’s digital age, despite the challenges of evolving tactics and understanding context and intent.
Publication date: 1 Dec 2023
Project Page: Not available
Paper: https://arxiv.org/pdf/2311.18730
