Integrating Argumentation Features for Enhanced Propaganda Detection in Arabic Narratives on the Israeli War on Gaza

Sara Nabhani, Claudia Borg, Kurt Micallef, Khalid Al Khatib

Abstract: Propaganda significantly shapes public opinion, especially in conflict-driven contexts like the Israeli-Palestinian conflict. This study explores the integration of argumentation features, such as claims, premises, and major claims, into machine learning models to enhance the detection of propaganda techniques in Arabic media. By leveraging datasets annotated with fine-grained propaganda techniques and employing crosslingual and multilingual NLP methods, along with GPT-4-based annotations, we demonstrate consistent performance improvements. A qualitative analysis of Arabic media narratives on the Israeli war on Gaza further reveals the model’s capability to identify diverse rhetorical strategies, offering insights into the dynamics of propaganda. These findings emphasize the potential of combining NLP with argumentation features to foster transparency and informed discourse in politically charged settings.