의료 담화의 한국어-아랍어 기계번역의 가용성과 한계

A Study on the Availability and Limitations of Korean-Arabic Machine Translation in Medical Discourse

This study explores the availability and limitations of AI-based translation tools in facilitating communication between Koreans and Arabs. Focusing on multilateral medical discourses involving three or more participants, the study compares ChatGPT’s translation outputs from 2024 and 2025 to assess improvements in quality. A parallel corpus was constructed using 4,313 words (17,074 characters) of Korean medical dialogues extracted from medical interpretation training materials, and translation errors were analyzed. The findings reveal significant enhancements in the 2025 translations, particularly in recognizing participants’ gender, understanding context, accurately translating medical terminology, and reflecting cultural elements. Notably, specifying the gender and role of medical staff and caregivers in prompts substantially reduced translation errors. Moreover, when prompts included “Arab Muslim patient,” culturally relevant Islamic expressions were naturally incorporated into the translations. However, persistent limitations remain, such as subject identification errors during speaker transitions. This study provides empirical evidence of the feasibility of AI-driven translation in multilateral medical dialogues and highlights the need for further research on more diverse scenarios and dialogues involving four or more participants.