본문 바로가기
  • Home

A Study of Translatability of Irony in ChatGPT

  • The Journal of Translation Studies
  • Abbr : JTS
  • 2023, 24(2), pp.131-160
  • DOI : 10.15749/jts.2023.24.2.005
  • Publisher : The Korean Association for Translation Studies
  • Research Area : Humanities > Interpretation and Translation Studies
  • Received : May 17, 2023
  • Accepted : June 22, 2023
  • Published : June 30, 2023

Soojung Park 1 Eunsil CHOI 1

1부산외국어대학교

Accredited

ABSTRACT

The advent of ChatGPT is making waves in various fields. The interactive AI can perform tasks commonly connected with human cognitive activities including playing games, analyzing medical data, and summarizing texts. In particular, the creativity of ChatGPT, which allows it to draw a painting or write a song, is a starting point for this study because translation is considered to be a creative work. Therefore, this study begins with the question of whether ChatGPT can translate a literary work, especially irony. Among the literary devices, irony is particularly difficult to translate because it requires a high level of complex perception and interpretation. Another reason why irony translation is difficult is that translators have difficulty choosing a translation strategy between the following two options: to literally translate irony to give readers a chance to find it, or to paraphrase irony to convey the intention of a writer. Until now, machine translation has had many difficulties not only in identifying irony, but also in translating it with an appropriate strategy. Therefore, this study aims to examine if ChatGPT can be utilized to translate irony. To answer the research question, we follow two processes. First, we take a source text and its target texts from human translation, machine translation using Google, and AI translation using ChatGPT, and then analyze them. The results show that ChatGPT understands and translates the meaning and intent of the source text more deeply than existing machine translation, even though the results were not as complete as the human translation. Secondly, we provide the AI with some information at the word, sentence, and context level so that it can understand cultural elements and contexts that include irony. In some cases, even when ChatGPT gets a clue, it does not understand or reflect the writer’s intention. However, it does actively utilize the provided information and the translation quality does improve. Therefore, our conclusion is three-fold. First, machine translation and AI translation can be used in different ways. Second, the interactive character of AI can be helpful to improve translation quality. And lastly, AI translation can be utilized to translate irony.

Citation status

* References for papers published after 2022 are currently being built.