Medical terminology issues: a feasibility study of machine translation in a low-resource language

Abstract

Medical knowledge may be targeted at a multitude of audiences, including researchers, health professionals, patients and the general public. For many reasons, like in other spheres, machine translation has gained its way into medical and clinical settings. Medical translation is complicated not only because of terminological issues, but also because of a variety of genres and text types ranging from clinical practice to education, research and dissemination. Adequacy and fluency of the translated text are all equally crucial. The aim of this study is to analyze machine translation output of medical texts with a focus on terminology issues in English to Lithuanian language pair. The results of this study on human evaluation of machine translated output reveal that the performance of machine translation systems is insufficient, and the raw output requires careful review. While the accuracy of terminology translation and adequacy of longer segments may be regarded fairly acceptable for non-professional uses, the fluency of the raw translation mostly produces poor readability. Our research corroborates the findings of other studies that machine translation may be employed as a complementary tool only and cannot be relied on as the only source, but its significance for medical professionals and the broad public cannot be ignored.