Arabic Named Entity Recognition: Approaches, Datasets, and Comparative Study

This survey delves into the challenges and advancements in Arabic Named Entity Recognition (NER), a cornerstone task in Natural Language Processing. Despite Arabic’s rich linguistic tapestry and the absence of comprehensive resources, significant strides have been made in NER methodologies, ranging from rule-based to cutting-edge deep learning techniques. We spotlight key datasets, including ‘ANERCorp’, ‘AQMAR’, and the fine grained ‘WikiFANE_Gold’ and ‘NewsFANE_Gold’. A notable highlight for a semi-supervised deep learning approach, which offers promising avenues for domain-specific applications.