Credibility Assessment for Arabic News on the Gaza War: A Hybrid Neural-Symbolic Pipeline

Abstract

While misinformation has long circulated online, the Gaza conflict has intensified its visibility and spread across news websites, online portals, and social media, complicating the credibility and long-term curation of conflict-related Arabic records, including historical accounts and written testimonies. This work proposes a hybrid framework for Arabic fake news detection that combines interpretable linguistic cues with contextual semantic representations. The approach integrates fuzzy logic-based handcrafted features capturing exaggerated and sensational linguistic patterns, AraBERT contextual embeddings for semantic understanding, and a CNN-based text feature extractor for local textual patterns. These complementary features are combined into a unified representation for downstream classification. Multiple machine learning and deep learning classifiers are evaluated to identify the most effective detection model. The resulting system is deployed as a real-time web browser plugin, enabling users to automatically assess the credibility of Arabic news content during browsing

Resources

Details