EUVIMEDEUVIMED entdecken
Crossref

A Study on Emotion Reproduction Patterns in Human and AI Translation Based on Sentiment Analysis: Focusing on the Chinese Translations of the Novel We Do Not Part

Hwayoung Song

Korea Association of Chinese Linguistics · 30.10.2026 · 126 · 475-502

DOI: 10.38068/kjcl.126.15 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Korea Association of Chinese Linguistics, online erschienen am 30.10.2026. Verfasst von Hwayoung Song.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study conducted a sentiment analysis of the Chinese translations of the novel We Do Not Part produced by a human translator, ChatGPT, and Claude, comparing patterns of emotion reproduction in human and AI translation. Through this analysis, the study examined, from multiple perspectives, the extent to which AI translation can reproduce the emotions embodied in human translation. To this end, a parallel corpus comprising the Chinese translations produced by the human translator, ChatGPT, and Claude was constructed, and quantitative analyses of emotion word usage, sentiment polarity distribution, emotion category distribution, and changes in emotion categories were conducted in RStudio using the sentiment lexicon developed by Dalian University of Technology. With regard to emotion word usage, Claude had the highest total token count and used a more diverse range of emotion words, whereas ChatGPT more closely resembled the human translation in both the frequency and use of emotion words. In terms of sentiment polarity distribution, positive sentiment predominated across all three translations; however, the human translation showed a higher proportion of positive sentiment, while both AI models displayed relatively higher proportions of negative sentiment. Regarding emotion category distribution, the categories of liking, dislike, and joy accounted for high proportions across all three translations, although the rankings of the lower-level emotion categories varied across the three translations. As for changes in emotion categories, the resulting curves showed broadly similar patterns overall, despite some differences in certain chapters. Moreover, ChatGPT and Claude tended to move in largely the same direction. The findings indicate that ChatGPT and Claude, on the whole, reproduce the emotions embodied in human translation with a considerable degree of similarity. It is hoped that this study will serve as a useful reference for understanding patterns of emotion reproduction in AI translation and for exploring effective collaboration between human translators and AI.

Zitieren

APA

Song, H. (2026). A Study on Emotion Reproduction Patterns in Human and AI Translation Based on Sentiment Analysis: Focusing on the Chinese Translations of the Novel We Do Not Part. Korea Association of Chinese Linguistics, 126, 475-502. https://doi.org/10.38068/kjcl.126.15

Vancouver

Song H. A Study on Emotion Reproduction Patterns in Human and AI Translation Based on Sentiment Analysis: Focusing on the Chinese Translations of the Novel We Do Not Part. Korea Association of Chinese Linguistics. 2026;126:475-502. doi:10.38068/kjcl.126.15

BibTeX
@article{Song2026,
  author = {Song, Hwayoung},
  title = {A Study on Emotion Reproduction Patterns in Human and AI Translation Based on Sentiment Analysis: Focusing on the Chinese Translations of the Novel We Do Not Part},
  journal = {Korea Association of Chinese Linguistics},
  year = {2026},
  volume = {126},
  pages = {475--502},
  doi = {10.38068/kjcl.126.15}
}
Kostenlos starten →