기계번역과 챗GPT의 번역생성문 비교 – 한일 화장품 홈페이지의 텍스트를 대상으로 –

Abstract

This paper analyzes whether machine translation and ChatGPT’s translation outputs reflect the linguistic findings in the field of contrastive studies of Korean and Japanese. For this purpose, I compare the features of translation outputs from Google translate, Papago translate, and ChatGPT using text from Korean and Japanese cosmetic websites. The results are summarized as follows. (I) Comparison of Korean-to-Japanese Translation Outputs Few instances where verbs occurred in the original Korean text were translated into Japanese verb ‘michibiku’ or ‘noun+case particle ending’, although both of them are preferred style in Japanese text. While Google translate mostly adapted Korean ‘-a/e cwu-ta’ verb literally into Japanese ‘te kureru’ using it’s corresponding form, Papago and ChatGPT translated it into simple verb, i.e without ‘te kureru’ in consideration of Japanese characteristics. (II) Comparison of Japanese-to-Korean Translation Outputs The examples where verbs appeared in the original Japanese text were translated into Korean ‘-a/e cwu-ta’ were often observed in Papago. Concerning the instances of ‘noun+case particle ending’, ChatGPT exclusively translated them into natural sentences in Korean by supplementing predicate absent in the original Japanese text.