A Comparative Linguistic Analysis Of Human And AI Medical Term Translations Into Arabic

Abstract

In recent years, artificial intelligence has become a dominant force in language translation, even in high-stakes fields such as medicine. Nevertheless, questions persist on the accuracy and context-specific validity of AI-provided translations of medical terms. This study aims to evaluate the quality of AI translations in comparison to those of human professionals, focusing on linguistic accuracy, clinical appropriateness, and adherence to medical discourse norms. It specifically analyzes the extent to which AI tools, such as ChatGPT-4, successfully transcribe English medical terminologies into Arabic, identifies recurrent linguistic difficulties faced in AI translation, and discuss possible avenues to enhance the quality of translation through AI. The study draws its conceptual foundation from Halliday's Ideational Functional Linguistics (SFL) approach and focuses on the ideational, interpersonal, and textual functions of medical language. The comparative qualitative analysis was undertaken using purposive sample selection of ten English medical terms from various clinical subfields. Each of these terms was translated by both AI and human translators and analyzed using a descriptive-analytical approach, taking into consideration term accuracy, syntagmatic structure, grammatical accuracy, and register appropriateness. The findings reveal that AI translations are structurally fluent but often lack semantic accuracy, subject-specific terminological usage, and register appropriateness. The comparison shows that human translations are invariably superior in terms of appropriateness to Arabic clinical conventions, particularly in procedural-to-diagnostic contexts and descriptions of pathophysiology. The current study concludes that AI translation tools demand substantial enhancement in accordance with exposure to specialized Arabic medical corpora, enhanced genre sensibility, and post-editing measures. These findings have significant implications for the integration of AI in healthcare communication and support the strategic objectives of Saudi Arabia’s Vision 2030 in advancing AI applications in medicine. The study contributes to the broader discourse on the responsible use of AI in sensitive domains, advocating for hybrid translation models that combine machine efficiency with human linguistic expertise.