<![CDATA[This study examined the performance of AI models in translating classical Arabic grammatical literature, focusing on Alfiyah Ibn Mālik and Naẓm al-Imrīṭī, two foundational texts marked by dense syntactic structures and strong pedagogical significance. ChatGPT and Gemini were evaluated in terms of translation accuracy, terminological precision, and contextual sensitivity. A panel of expert evaluators with more than fifteen years of experience in Arabic instruction assessed each model’s capacity to apply syntactic rules, preserve semantic coherence, and maintain stylistic and didactic integrity. The aim and scope of the paper centred on measuring translation quality through a combined framework of METEOR scoring and human expert judgement. Qualitative evaluation further explored the models’ adaptability to classical Arabic rhetorical patterns and instructional conventions. The results showed that ChatGPT achieved higher lexical alignment and word-level accuracy than Gemini according to METEOR scores; however, both models demonstrated notable limitations in rendering idiomatic expressions and conveying deeper grammatical and contextual meanings. Statistical analysis using the Mann–Whitney U test revealed no significant difference between the two models, underscoring the limited explanatory power of automated metrics when applied to highly structured classical texts. These findings underscored the ongoing need for expert validation beyond numerical scoring and supported the adoption of a hybrid translation framework, in which AI-generated outputs are systematically refined through scholarly review. Future research was suggested to broaden the textual corpus, incorporate additional AI models and evaluation metrics, and further strengthen expert-led validation to enhance the reliability of AI-assisted translation in advanced Arabic grammatical studies.]]>
Copyrights © 2026
