본문 바로가기
  • Home

AI-based translation of Turkish discourse markers into Korean: A study of dialogues in “Yaşar Ne Yaşar Ne Yaşamaz”

  • The Journal of Translation Studies
  • Abbr : JTS
  • 2026, 27(1), pp.319~353
  • DOI : 10.15749/jts.2026.27.1.010
  • Publisher : The Korean Association for Translation Studies
  • Research Area : Humanities > Interpretation and Translation Studies
  • Received : February 14, 2026
  • Accepted : March 16, 2026
  • Published : March 31, 2026

F. H. Kubra Aydin 1 Yu, Eunmi 2

1한국외국어대학교
2앙카라대학교 한국어문학과

Accredited

ABSTRACT

This study examines how Turkish discourse markers are translated into Korean in literary texts, with a focus on the pragmatic processing capabilities of AI-based machine translation systems. Using the Turkish novel “Yaşar Ne Yaşar Ne Yaşamaz” as the primary source, three high-frequency discourse markers—işte, yani, and hani—were analyzed across four target texts: a published human translation, Google Translate output, ChatGPT zero-shot output, and ChatGPT output with an RGC (Role-Goal-Context) prompt. Each translation was assessed in terms of pragmatic equivalence and classified into four categories: discourse marker, non-discourse marker, omission, and mistranslation. The findings reveal that discourse markers were not systematically omitted but were frequently restructured into non-discourse-marker forms such as sentence-final endings, particles, and adverbs in Korean. ChatGPT demonstrated notably higher pragmatic equivalence than Google Translate, with accuracy reaching 97.8% under RGC prompt conditions, compared to 77.8% for Google Translate. These results indicate that LLM-based translation systems are better able to interpret discourse markers in context and that prompt design can substantially enhance translation performance. This study contributes to emerging discussions on AI-assisted literary translation, particularly regarding the pragmatic handling of discourse markers in the underexplored Turkish-Korean language pair.

Citation status

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