기계번역의 문학번역 적용에 관한 예비 연구: 2025년 챗GPT와 NMT 결과물을 중심으로

초록

Recent advances in machine translation have expanded its perceived applicability beyond informational texts to the domain of literary translation, particularly through the use of large language models with prompt-based interaction. However, literary translation differs fundamentally from general translation in that it requires the recreation of aesthetic form, cultural context, and narrative voice rather than the mere transfer of propositional meaning. Accordingly, this study investigates the extent to which neural machine translation (NMT) and ChatGPT-based translation have developed in comparison with earlier machine translation outputs. In addition, it examines whether ChatGPT-based translation outperforms NMT in literary translation tasks through automatic evaluation, expert assessment, and quantitative error analysis. The results indicate that ChatGPT demonstrates overall improvements in translation quality and stronger contextual understanding, suggesting its potential usefulness as a supportive tool in literary translation. Nevertheless, persistent limitations remain, including omissions, overgeneration, lexical inconsistency, inadequate handling of out-of-vocabulary items, and difficulties in conveying implicit meanings and maintaining global narrative coherence. These findings indicate that, despite recent technological advances, machine translation should be regarded as a supplementary aid rather than a substitute for human literary translators.

키워드