Advancements and challenges in Arabic question answering systems: a comprehensive survey

This paper explores the development and challenges of Arabic Question Answering Systems (QAS), with a particular focus on addressing the linguistic nuances and data scarcity unique to Arabic. It categorizes Arabic QAS by domain (open or closed), question type (factoid, causal, complex), and modeling approach (rule-based, machine learning, deep learning). Key recent systems are reviewed, including their methodologies, datasets, and performance metrics, highlighting the growing role of deep learning techniques. State-of-the-art NLP methods, especially transformers, have greatly advanced QAS, although these gains are more substantial for English. Despite recent progress, Arabic QAS research remains limited by resource constraints, including a reliance on translated datasets and the absence of comprehensive Arabic benchmarks.