AI vs Human Translation of Oceanography Discourse (English to Arabic)

Abstract

Translating general technico-scientific terminology is a challenging task that becomes more so when handling oceanography discourse, particularly from English into Arabic. With recent developments in artificial intelligence (AI) and its translation applications, the question becomes whether AI tools can produce human-like translations of such specialized terminology. This study examines human and AI translations of 150 oceanography terms from English into Arabic yielding a corpus of 450 translations. The terms were mined from Guide to the Oceans (Pernetta, 2004) for which the human translation (HT) into Arabic was carried out by Fathi Majdalawieh titled دليل المحيطات (Pernetta, 2013). AI-translations were generated by ChatGPT (GPT) and Google Gemini (GG). Using Pedersen’s (2011) taxonomy of translation strategies, the Arabic renderings were interpretatively analyzed. The analysis indicates HT and GPT employed primarily the substitution strategy (64% and 59.33%, respectively, compared to 20% by GG). GG topped the three in its use of direct translation (58%), compared to GPT (24%) and HT (22.67%). Such findings suggest that GPT’s output was closer to HT’s in maintaining meanings and nuances of oceanography terminology. The findings offer insights into the current capabilities of AI models in translating technico-scientific terminology, with GPT showing a closer alignment with HT than GG.