본문 바로가기
  • Home
Basic Search

An analysis of errors in ChatGPT’s Korean-Chinese interpretation

  • The Journal of Translation Studies
  • Abbr : JTS
  • 2025, 26(2), pp.139~167
  • DOI : 10.15749/jts.2025.26.2.005
  • Publisher : The Korean Association for Translation Studies
  • Research Area : Humanities > Interpretation and Translation Studies
  • Received : May 15, 2025
  • Accepted : June 15, 2025
  • Published : June 30, 2025

Su, Kong 1

1이화여자대학교

Accredited

ABSTRACT

This study investigates the types and causes of errors in Korean-Chinese interpretation outputs generated by ChatGPT. Using 60 minutes of recorded speeches and interviews, the study categorizes errors into three domains—fidelity, fluency, and pragmatics—each with its own set of subcategories. The findings reveal that accuracy-related errors—such as content distortion, omissions, grammatical mistakes, and misuse of technical terms—were frequent. Fluency-related and pragmatic errors, including awkward expressions, mispronunciations, and contextually inappropriate usage, were also observed, especially in cases involving Sino-Korean words, neologisms, unclear audio, or dialects. Most errors occurred only in speech-based interpretation, not in text-based translation of the same source, suggesting that real-time processing constraints and speech recognition issues are key factors. Technical limitations, such as fragmented output and interference from background noise, were also noted. Although honorific misuse and content addition were defined as error types, they were not detected, likely due to time constraints and limited data. The study concludes that while ChatGPT performs well overall, human review remains essential.

Statistics

  • 6 Viewed
  • 4 Downloaded

Tools

Issue List

Citation status

* References for papers published after 2023 are currently being built.

Basic Search Advanced Search