한국어–아랍어 의료 안내문 번역에서 GPT-4o의 구조적⋅화용적 적절성 분석과 교육적 함의

Abstract

This study investigates the structural and pragmatic adequacy of GPT-4o in translating Korean medical instruction leaflets into Arabic, with an emphasis on its implications for medical interpreter training. A parallel corpus of 122 sentences was analyzed through both quantitative and qualitative methods, comparing GPT-4o outputs with professional human translations. Results show that GPT-4o demonstrates reasonable competence in structural aspects such as sentence clarity, information sequencing, and occasionally even producing more concise and straightforward expressions than human translators. However, clear limitations emerged in pragmatic dimensions. GPT-4o frequently failed to reproduce politeness strategies, indirectness, and cultural sensitivity, which are essential in Arabic medical discourse. In particular, expressions that ignored religious and cultural contexts risked undermining the reliability and acceptability of medical instructions for Arabic-speaking patients. Such findings underscore that while GPT-4o may serve as a useful supplementary tool, it cannot substitute for the nuanced pragmatic and cultural adjustments that human translators provide. Drawing on these results, this study proposes several pedagogical applications. First, GPT-4o translations can be utilized as practice material for post-editing activities, helping learners strengthen skills in error detection, stylistic adjustment, and structural refinement. Second, comparative analysis between human and machine translations can enhance learners’ critical awareness of machine translation limitations and the necessity of human intervention. Third, training activities focusing on cultural sensitivity, taboo avoidance, and politeness strategies are vital for preparing interpreters to function effectively in multicultural healthcare contexts. In conclusion, GPT-4o demonstrates partial potential in handling structural aspects of Korean–Arabic medical translation, but pragmatic refinement and cultural adaptation remain indispensable domains of human expertise. Future research should expand to diverse medical genres and clinical contexts, incorporating empirical validation of patient reception, and extend comparative analysis across multiple languages to more precisely assess the role of AI-assisted translation in medical interpreter education.