Sentiscan: Leveraging Multilingual NLP, RL, and XAI for Robust Governmental Intelligence Using Sentiment Analysis with Arabic Dialect Focus
Haytham Tarek Mohammed Fetooh, Eman Ali Metwally · 2025
As government organizations increasingly rely on social media analysis to gauge public opinion, the need for robust monitoring tools is paramount. This paper presents SentiScan, a sentiment analysis platform designed for governmental use. SentiScan integrates multilingual NLP with specialized handling of Arabic dialects, an intelligent multimodel ensemble optimized by reinforcement learning (RL), and a novel data fusion component. The architecture incorporates explainable AI (XAI) for transparency and an adaptive RL feedback loop for continuous improvement. Comparative analysis demonstrates SentiScan's novel contributions in model architecture, language coverage, and adaptability. Experimental results show competitive performance and the generation of interpretable insights valuable for policy evaluation, crisis management, and strategic communications.