Ali Mohammed Aloqla

Arabic question-answering systems in high-sensitivity domains require not only accurate retrieval but also reliable evidence grounding and effective hallucination mitigation, as incorrect or unsupported responses may have serious consequences. Existing retrieval and generation approaches do not fully integrate reliable lexical retrieval, semantic understanding, and evidence-based answer verification within a unified framework for these domains. To address this limitation, this study proposes an Enhanced Hybrid Retrieval-Augmented model that combines BM25-based lexical retrieval, dense semantic scoring, semi-structured metadata, domain-aware classification, source-trust and freshness indicators, and evidence-based answer verification. The model was evaluated on a corpus of 128,297 Arabic documents using 200 expert-validated questions and compared against BM25, Dense, and Classical Hybrid retrieval configurations. Expert assessment showed that the Enhanced Hybrid Model produced 198 fully grounded correct answers out of 200, achieving a grounded-correctness rate of 99.0%, compared with 96.5% for Classical Hybrid, 87.0% for BM25, and 86.5% for Dense. It also achieved the strongest answer-level Hit@5 performance, the best golden-answer ordering, and the highest dynamic mixture-based F1-score. These findings demonstrate that integrating reliable retrieval, metadata-aware ranking, and evidence-grounded answer verification can substantially improve the reliability and trustworthiness of Arabic question-answering systems in high-sensitivity domains, providing a practical foundation for future evidence-based intelligent information systems.

Question-Answering (QA) has become a pivotal topic in Natural Language Processing (NLP), facilitating machines' comprehension and response to human inquiries in natural language. Although QA systems for English and other high-resource languages have been extensively studied, Arabic QA remains under-investigated and faces several linguistic and technical challenges. This paper offers an extensive analysis of deep learning-based Arabic QA systems, emphasizing extractive, generative, and hybrid architectures. This study analyzes the fundamental issues in Arabic processing, outlines essential datasets, and provides a classification of QA methodologies. Furthermore, it identifies several research gaps, including the absence of domain-specific models, limited generative question answering, and insufficient use of retrieval-augmented architectures. To overcome these deficiencies, a Fatwa-based dataset, currently under development, can serve as a resource for future research on domain-specific Arabic QA. This study also delineates prospective trajectories, emphasizing the promise of Retrieval-Augmented Generation (RAG), few-shot learning, and dialect-aware models in propelling the discipline forward.