본문 바로가기
  • Home

A Study on the Improvement of Contextual Interpretation in Korean-Japanese Machine Translation —Focusing on Homonyms—

  • 日本硏究
  • 2026, (64), pp.29~48
  • Publisher : The Center for Japanese Studies
  • Research Area : Humanities > Japanese Language and Literature
  • Received : January 11, 2026
  • Accepted : January 26, 2026
  • Published : February 20, 2026

Hansik Kim 1

1한국외국어대학교

Accredited

ABSTRACT

This study investigatedsexamined whether and how the accuracy of AI-based machine translation has improved, with particular attention paidgiven to contextual interpretation, a long-standing weakness of machine translation systems. From the 305 homonyms examined analyzed byin Kim(2022), 116 lexical items with high translation difficulty that had previously resulted in mistranslations were selected. Newspaper articles containing these items were retranslated in October 2025, and the translation outputs from two different time points were comparatively compared analyzed. The translations produced by ChatGPT, a generative AI system publicly released in 2023, were also included in the analysis. The results indicated showed that Google Translate showed demonstrated only a slight improvement in accuracy, whereas Papago exhibited experienced a small decline, suggesting that the homonym translation accuracy in these systems didhas not improve substantially improved. In contrast, ChatGPT achieved a markedly higher score, clearly outperforming surpassing both Google Translate and Papago. These findings indicatesuggestdemonstrate a significant improvement in the accuracy of AI-based MTAI-based machine translation accuracy. Moreover, ChatGPT not only translates homonyms accurately in context but also produces fluent and highly readable translations across the entire texts.

Citation status

* References for papers published after 2024 are currently being built.