生成式AI中英文化负载词翻译策略与质量研究——以《红楼梦》ChatGPT与DeepSeek译本为例
A Study on the Translation Strategies and Quality of Culture-Loaded Terms in Generative AI—A Comparative Analysis of ChatGPT and DeepSeek Translations of The Story of the Stone
摘要: 现有研究多聚焦于通用翻译质量评估,对文化负载词翻译机制关注不足,本研究深入探讨了生成式AI在文化负载词翻译中的策略分化与语境作用问题。以《红楼梦》前八十回五类文化负载词为对象,基于Nida文化五分法与Venuti归化–异化理论,采用定量与定性相结合方法,对比分析GPT-4o与DeepSeek的翻译表现。研究发现,GPT-4o系统性倾向归化策略,与霍克斯译本匹配度更高;DeepSeek倾向异化策略,与杨宪益译本契合度更优。词类翻译质量差异显著,物质文化词表现最佳,语言与社会文化词最差。值得关注的是,语境对翻译质量无正向增益,反而可能干扰模型对文化词的聚焦,其干扰强度与文化图式依赖性相关。本研究揭示了生成式AI作为文化中介的潜在立场,为AI助力中国古典文学海外传播提供了重要的实证参考。
Abstract: While existing research has largely focused on general translation quality assessment, there remains insufficient attention to the translation mechanisms of culture-loaded terms. This study delves into the strategic divergence and contextual influence of generative AI in translating such terms. Taking five categories of culture-loaded terms from the first eighty chapters of The Story of the Stone as the research object, and drawing upon Nida’s five-category cultural framework and Venuti’s theory of domestication and foreignization, this study employs a mixed-methods approach to compare the translation performance of GPT-4o and DeepSeek. Findings indicate that GPT-4o systematically leans toward domestication, aligning more closely with the Hawkes translation, whereas DeepSeek exhibits a preference for foreignization, showing greater affinity with the Yang Xianyi translation. Translation quality varies significantly across term categories, with material culture terms achieving the highest accuracy and linguistic and social culture terms the lowest. Notably, contextual information does not contribute positively to translation quality; rather, it may distract the model from focusing on culture-specific terms, with the degree of interference correlating with the dependency on cultural schemata. This study reveals the implicit stance of generative AI as a cultural mediator and provides valuable empirical insights for leveraging AI in the global dissemination of classical Chinese literature.
文章引用:王禹雯. 生成式AI中英文化负载词翻译策略与质量研究——以《红楼梦》ChatGPT与DeepSeek译本为例[J]. 现代语言学, 2026, 14(1): 97-106. https://doi.org/10.12677/ml.2026.141015

参考文献

[1] Vaswani, A., et al. (2017) Attention Is All You Need. Proceedings of the 31st International Conference on Neural Information Processing Systems, Long Beach, 4-9 December 2017, 6000-6010.
[2] OpenAI (2023) GPT-4 Technical Report.
[3] DeepSeek (2024) DeepSeek LLM Technical Report.
[4] 冯志伟, 张灯柯. 语言模型与人工智能[J]. 外语研究, 2024, 41(1): 1-19, 112.
[5] Bender, E.M., Gebru, T., McMillan-Major, A. and Shmitchell, S. (2021) On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 3-10 March 2021, 610-623. [Google Scholar] [CrossRef
[6] 李德凤, 王华树, 刘世界. 国家翻译技术能力与大语言模型[J]. 上海翻译(中英文), 2025(2): 18-24.
[7] 文旭, 田亚灵. ChatGPT应用于中国特色话语翻译的有效性研究[J]. 上海翻译, 2024(2): 27-34.
[8] 毛文伟, 朱海莹. 计量文体学视角下的汉日机器翻译语言特征研究——以《阿Q正传》日译本为例[J]. 日语学习与研究, 2024(6): 63-73.
[9] Deng, Y. and Sun, H. (2023) Cultural Information Transfer in Neural Machine Translation: A Systematic Evaluation of Large Language Models. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics, Toronto, 9-14 July 2023, 1250-1265.
[10] Nida, E.A. (1964) Toward a Science of Translating. Brill. [Google Scholar] [CrossRef
[11] Venuti, L. (1995) The Translator’s Invisibility: A History of Translation. Routledge.
[12] Fisher, R.A. (1936) Statistical Methods for Research Workers. 6th Edition, Oliver and Boyd.
[13] Papineni, K., Roukos, S., Ward, T. and Zhu, W. (2001) BLEU. Proceedings of the 40th Annual Meeting on Association for Computational Linguistics-ACL‘02, Philadelphia, 7-12 July 2002, 311-318. [Google Scholar] [CrossRef
[14] 许国璋. 文化负载词与英语语言教学[J]. 现代外语, 1980(4): 21-27.
[15] Hartmann, R.R.K. and James, G. (2000) Dictionary of Lexicography. Routledge, 3.
[16] Sperber, D. and Wilson, D. (1986) Relevance: Communication and Cognition. Blackwell.
[17] Rosch, E. (2024) Principles of Categorization. In: Rosch, E. and Lloyd, B.B., Eds., Cognition and Categorization, Routledge, 27-48. [Google Scholar] [CrossRef
[18] Rumelhart, D.E. (2017) Schemata: The Building Blocks of Cognition. In: Spiro, R.J., Bruce, B.C. and Brewer, W.F., Eds., Theoretical Issues in Reading Comprehension, Routledge, 33-58. [Google Scholar] [CrossRef