Earticle

초록

영어

With the rapid development of generative AI, ChatGPT has been increasingly applied to translation tasks. However, translating classical poetry remains challenging because it demands both semantic accuracy and stylistic reproduction. This study investigates ChatGPT’s performance in translating elegiac poems in Romance of the Three Kingdoms. Using human translation as a reference, it compares outputs generated under literal and poetry-oriented prompts. A multidimensional framework combining automated and human evaluation is adopted. Cosine similarity, BERTScore, and chrF2 are used to assess semantic and formal alignment, while human evaluators rate fidelity, metaphorical expression, style, and overall acceptability. The findings indicate that poetry-oriented prompts improve the translations to a certain extent, but still fail to reproduce the subtle expressiveness achieved by human translation. The study proposes a practical framework for evaluating small-sample classical poetry translation and offers empirical insight into ChatGPT’s applicability to this task.

목차

Abstract
1. 引言
2. 先行研究回顾
3. 研究对象与方法
3.1 研究文本
3.2 评价方法
4. 评价结果分析
4.1 自动化评价结果分析
4.2 问卷测评结果分析
5. 结论
参考文献

저자

• 李震宁 [ 이진저 | 对外经济贸易大学/ 副教授 ] 第一作者
• 张珣婕 [ 장순첩 | 高丽大学国语国文系/ 博士 ] 通讯作者
• 王梦 [ 왕몽 | 北京外国语大学/ 博士后；哈尔滨师范大学讲师 ] 第三作者

참고문헌