当前位置:人文与社会科学学刊 > 2026年11月-2卷11期 > 文章详情
Open Access Crossmark - Check for updates
PDF 本期PDF

基于DeepSeek的农业文本翻译提示词研究

A Study on the Impact of Prompt Engineering on the Translation Quality of Agricultural Texts: A Case Study of DeepSeek

作者:徐月晨,李平
单位:南京农业大学,南京农业大学
卷期:2026年11月-2卷11期
通讯作者:李平
页码:227-233
发布时间:2026-11-10
总浏览量:260

摘要

随着中国农业科技“走出去”战略的推进,农业文本外译质量日益成为关键问题。本文以《中国的粮食安全》白皮书为语料、官方译文为参考,采用DeepSeek模型,比较四种提示词策略下的译文效果。研究发现,提示词工程能够明显提升农业文本翻译质量。其中,“角色扮演+少样本”策略在术语准确性和文体规范性方面表现最优,“角色扮演+思维链”策略有助于逻辑显化,但也存在术语创新风险。研究表明,提示词工程可为专业文本翻译提供有效优化路径。

关键词

提示词工程;大语言模型;农业文本;翻译

Abstract

With China's agricultural science and technology increasingly "going global," the quality of agricultural translation has become a key issue. Taking Food Security in China as the corpus and its official English version as the reference, this study uses DeepSeek to compare the translation performance of four prompt strategies. The results show that prompt engineering can improve translation quality. The "role-playing + few-shot" strategy performs best in terminological accuracy and stylistic consistency, whereas the "role-playing + chain-of-thought" strategy strengthens logical explication but may increase the risk of terminological innovation. The findings suggest that prompt engineering provides a practical pathway for optimizing professional text translation in agricultural contexts.

Keywords

prompt engineering; large language models; agricultural texts; translation

引用本文

徐月晨,李平. 基于DeepSeek的农业文本翻译提示词研究[J]. 人文与社会科学学刊. 2026, 2 (11): 227-233. DOI: 10.70693/202610160468.

APA引用

徐月晨, 李平. (2026). 基于DeepSeek的农业文本翻译提示词研究. 人文与社会科学学刊, 2 (11), 227-233. https://doi.org/10.70693/202610160468

参考文献

[1] 石琨,王植,时朝. 我国农业科技“走出去”的实践与思考[J]. 中国农业科技导报,2024,26(11):1-6.
[2] 刘玉琪,王珍. 傅兰雅读者意识对中国农业术语外译的启示[J]. 中国科技术语,2025,27(05):65-67.
[3] 张雨薇,田牧野. 治理剧场视野下中国农技推广海外实践的创新机制研究——基于中坦农技合作项目的田野观察[J]. 中国科技论坛,2023,(11):179-188.
[4] 樊景超.大模型驱动水稻种质文本挖掘提速[N/OL]. 中国科学报,2025.
[5] Jiao, W. et al. Is ChatGPT a Good Translator? Yes with GPT-4 as the Engine[EB/OL]. 2023. arXiv:2301.08745.
[6] Robinson, N. R. et al. ChatGPT MT: Competitive for High- (but not Low-) Resource Languages[EB/OL]. 2023. arXiv:2309.07423.
[7] 赵衍、张慧、杨祎辰.大语言模型在文本翻译中的质量比较研究——以《繁花》翻译为例[J].外语电化教学,2024,(04):60-66+109.
[8] 张曙康、赵朝永.大语言模型之于文学翻译的适切性研究——基于多指标评估的《边城》多模型译文质量对比[J].中国翻译,2025,22(04):85-95.
[9] 杨锋昌. ChatGPT对译员的思考与启示——以越南语法律翻译为例[J].中国科技翻译,2023,36(03):27-30+4.
[10] 王和松、马柯昕.人工智能翻译应用的对比研究——以生物医学文本为例[J].中国科技翻译,2023,36(03):23-26.
[11] 余静. DeepSeek翻译能力探索——以文学文本与金融类文本为例[J].中国翻译,2025,46(03):172-179.
[12] Gao, Y. et al. How to Design Translation Prompts for ChatGPT: An Empirical Study[EB/OL]. 2023. arXiv:2304.02182.
[13] Brown, T. B. et al. Language models are few-shot learners[A]. Advances in Neural Information Processing Systems[C]. 2020, 33: 1877-1901.
[14] Wei, J. et al. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models[A]. Advances in Neural Information Processing Systems[C]. 2022.
[15] White, J. et al. A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT[A]. Proceedings of the 56th Hawaii International Conference on System Sciences[C]. 2023.
[16] Reynolds, L., & McDonell, K. Prompt Programming for Large Language Models: Beyond the Few-Shot Paradigm. In Extended Abstracts of the 2021 CHI Conference on Human Factors in Computing Systems (pp.1–7). Association for Computing Machinery. 2021.
CC BY 本文依据 知识共享署名 4.0 国际许可协议 授权发布。允许他人在署名原作者及来源的前提下复制、传播和使用本文。

编辑部微信

回到顶部