Translation of Medical Terms by AI: A Comparative Linguistic Study of Microsoft Copilot and Google Translate

This study aims to evaluate the accuracy of translation equivalents to medical terms given by Microsoft Copilot (MC) and Google Translate (GT), and their semantic, contextual, and syntactic deficiencies in translation. A purposive, random sample of 204 English and Arabic medical terms was translated from English to Arabic and Arabic to English by MC and GT. Results showed that MC and GT gave accurate equivalents to 68.6% and 74.5% of the medical terms respectively. Both gave more correct equivalents to Arabic than English terms. For example, both gave correct equivalents to الدهون الثلاثية “triglycerides” and could recognize that العناية المركزة/الفائقة/الحثيثة have the same English equivalent “intensive care”. Nevertheless, there are semantic, contextual, and syntactic inaccuracies in MC and GT translations. For instance, GT failed to translate some terms correctly. It transliterated ابو دغيم “Abu Dhghaim”, and translated العشى الليلي (Night Dinner), الذئبة الحمراء (red wolf) literally. Both gave compound equivalents with different word orders. For “fibroglandular tissues”, MC yielded الليفية الغديّة الأنسجة and GT yielded الأنسجة الغدية الليفية. Both made errors in definiteness as الحماض الكيتوني by GT. MC gave an explanatory equivalent for “diabetic ketoacidosis “حمضية الدم المتأتية من السكري, whereas GT gave a concise equivalent (الحماض الكيتوني السكري). MC gave the same equivalent ‘lupus’ for الثعلبة &الذئبة الحمراء and the extraneous translation الغدي الليفية with awkward grammatical agreement and derivation. Similarly, GT gave تعظم ليفي which is semantically inaccurate. Both made contextual errors as approach/method for نهجة. Recommendations for translation pedagogy are given.