图书情报工作

图书情报工作 >

2026 , Vol. 70 >Issue 17: 34 - 49

DOI: https://doi.org/10.13266/j.issn.0252-3116.2026.17.003

专题：大语言模型支持下的文化与典籍计算人文研究

领域大语言模型下的古诗词古白翻译应用研究

  • 叶文豪 ,
  • 赵文华 ,
  • 王希羽 ,
  • 刘浏
展开
  • 1 南京农业大学信息管理学院 南京 210095;
    2 人文与社会计算江苏省高校哲学社会科学重点研究基地 南京 210095
叶文豪，博士，博士后；赵文华，硕士研究生；王希羽，硕士研究生；刘浏，副教授，博士，硕士生导师，通信作者，E-mail：liuliu@njau.edu.cn。

收稿日期: 2025-05-07

  修回日期: 2025-08-05

  网络出版日期: 2026-08-20

基金资助

本文系国家社会科学基金重大项目“中国古代典籍跨语言知识库构建及应用研究”（项目编号：21&ZD331）、江苏省社会科学基金一般项目“《二十四史》人物知识图谱构建与应用研究”（项目编号：24TQB004）和江苏省社会科学基金青年项目“先秦典籍外译知识库构建及应用研究”（项目编号：23TQC004）研究成果之一。

收起

Research on the Application of Classical Poetry and Classical-to-Vernacular Translation Under Domain-Specific Large Language Models

  • Ye Wenhao ,
  • Zhao Wenhua ,
  • Wang Xiyu ,
  • Liu Liu
Expand
  • 1 College of Information Management, Nanjing Agricultural University, Nanjing 210095;
    2 Research Center for Humanities and Social computing of Nanjing Agricultural University, Nanjing 210095
Ye Wenhao, PhD, Postdoc; Zhao Wenhua, Master’s Candidate; Wang Xiyu, Master’s Candidate; Liu Liu, Associate Professor, PhD, Master’s Supervisor, Corresponding Author, E-mail: liuliu@njau.edu.cn.

Received date: 2025-05-07

  Revised date: 2025-08-05

  Online published: 2026-08-20

Supported by

This work is supported by Major Program of National Social Science Fund of China, titled “Research on the construction and application of cross-language support library for ancient Chinese classics” (Grant No. 21&ZD331); the general project of the Jiangsu Provincial Social Science Fund, titled “Research on the Construction and Application of Knowledge Graphs for Figures in the Twenty-Four Histories” (Grant No. 24TQB004); and the youth project of the Jiangsu Provincial Social Science Fund, titled “Construction and Application of the Knowledge Base for the Translation of Pre-Qin Classics” (Grant No. 23TQC004).

Fold

摘要

[目的/意义]大语言模型为古诗词机器翻译带来了直接助力。通过考察和对比大语言模型在古诗词翻译中的表现，尤其关注垂直领域大模型、训练集规模和朝代对翻译效果的影响，以期为未来古诗词翻译模型的优化提供参考。[方法/过程]使用整体和分类两种方法构建指令微调数据集，以BLEU、ROUGE作为评估指标并结合智能体（Agent）进行评估，选择经过古诗词数据预训练与微调的Qwen1.5-7B-poem模型与Qwen、Baichuan系列大语言模型。通过比较各模型在不同规模训练集下的翻译效果，研究分析大模型对语言特征和翻译规则的学习能力。此外，对比不同朝代诗词的翻译效果，以评估古诗词风格对翻译质量的影响。[结果/结论]结果表明，随着微调数据规模的增加，各模型的翻译准确性和流畅度都得到显著提高，表现出其对古诗词的语言特征与翻译规则的学习能力，符合大语言模型的规模效应。Qwen1.5-7B-poem表现最为突出，表明垂直领域大模型在领域性任务中更能发挥规模效应。此外，不同朝代的诗词翻译效果存在显著差异，表明语言的时代特点以及诗词的时代风格对模型翻译性能具有影响。大语言模型在古诗词翻译中的表现受训练数据集规模和朝代等多种因素的影响，未来研究可以进一步优化这些因素，以提升模型的泛化能力和翻译质量。

关键词： 大语言模型; 机器翻译; 古诗词; 指令微调

本文引用格式

叶文豪 , 赵文华 , 王希羽 , 刘浏 . 领域大语言模型下的古诗词古白翻译应用研究[J]. 图书情报工作, 2026 , 70(17) : 34 -49 . DOI: 10.13266/j.issn.0252-3116.2026.17.003

Abstract

[Purpose/Significance] Large language models provide direct assistance for the machine translation of classical Chinese poetry. This paper aims to examine and compare the performance of large language models in translating classical Chinese poetry, with a particular focus on the impacts of domain-specific models, training dataset sizes, and dynastic periods on translation effectiveness, in order to provide insights for the future optimization of classical poetry translation models. [Method/Process] The study used both holistic and classification methods to construct instruction-tuning datasets, evaluating the models using metrics such as BLEU and ROUGE, as well as agent-based assessments. The models compared include the Qwen1.5-7B-poem, pre-trained and fine-tuned on classical poetry data, as well as the Qwen and Baichuan series large language models. By comparing the translation performance of various models across different dataset sizes, the study analyzed the models' ability to learn linguistic features and translation rules. Additionally, the research compared the translation effectiveness of poems from different dynasties to evaluate the influence of poetic styles and linguistic features on translation quality. [Result/Conclusion] The results show that as the size of the fine-tuning data increases, the translation accuracy and fluency of all models improve significantly, demonstrating their ability to learn the linguistic features and translation rules of classical Chinese poetry, consistent with the scale effect of large language models. The Qwen1.5-7B-poem model performes the best, indicating that specialized large models achieve better scale effects in domain-specific tasks. Moreover, there were significant differences in the translation quality across poems from different dynasties, high-lighting the impact of factors such as temporal linguistic features and poetic styles on model performance. Overall, the performance of large language models in classical Chinese poetry translation is influenced by multiple factors, including the training dataset size and dynastic periods. Future research can further optimize these factors to improve the models' generalization and translation quality.

Key words： large language models; machine translation; classical Chinese poetry; instruction tuning

参考文献

[1] 文军.汉语古诗英译的描写模式研究——以杜甫诗歌英译的个案为例[J]. 外国语(上海外国语大学学报), 2013, 36(5):72-81.(WEN J. Research on the descriptive model of English translations of classical Chinese poems:a case study on Du Fu[J]. Journal of Foreign Languages(Shanghai International Studies University), 2013, 36(5):72-81.)
[2] 蒋锐滢,崔磊,何晶,等.基于主题模型和统计机器翻译方法的中文格律诗自动生成[J]. 计算机学报, 2015, 38(12):2426-2436.(JIANG R Y, CUI L, HE J, et al. Topicmodel and statistical machine translation based computer assisted poetry generation[J]. Chinese journal of computers, 2015, 38(12):2426-2436.)
[3] 朱丹浩,赵志枭,胡蝶,等.领域大语言模型下的古籍词性标注应用研究[J]. 科技情报研究, 2024, 6(2):21-29.(ZHU D H, ZHAO Z X, HU D, et al. Research on the application of part-of-speech tagging of ancient books under the domain large language model[J]. Scientific information research, 2024, 6(2):21-29.)
[4] 朱丹浩,赵志枭,吴娜,等.基于领域大语言模型的古籍分词研究[J]. 科技情报研究, 2024, 6(2):11-20.(ZHU D H, ZHAO Z X, WU N, et al. Research on word segmentation of ancient books based on domain large language model[J]. Scientific information research, 2024, 6(2):11-20.)
[5] ICHIEN N, STAMENKOVIC D, HOLYOAK K J. Large language model displays emergent ability to interpret novel literary metaphors[J]. Metaphor and symbol, 2024, 39(4):296-309.
[6] ZHAO W X, ZHOU K, LI J, et al. A survey of large language models[PP]. arXiv preprint arXiv:2303.18223, 2023.
[7] DEVLIN J. Bert:pre-training of deep bidirectional transformers for language understanding[PP]. arXiv preprint arXiv:1810.04805, 2018.
[8] KAPLAN J , MCCANDLISH S , HENIGHAN T, e t a l.Scaling laws for neural language models[PP]. arXiv preprint arXiv:2001.08361, 2020.
[9] WEI J, TAY Y, BOMMASANI R, et al. Emergent abilities of large language models[PP]. arXiv preprint arXiv:2206.07682,2022.
[10] 张辉,刘鹏,姜钧译,等. ChatGPT:从技术创新到范式革命[J]. 科学学研究, 2023, 41(12):2113-2121.(ZHANG H, LIU P, JIANG J Y, et al. ChatGPT:from technological innovation to paradigm revolution[J]. Studies in science of science, 2023,41(12):2113-2121.)
[11] THOPPILAN R, DE FREITAS D, HALL J, et al. Lamda:language models for dialog applications[PP]. arXiv preprint arXiv:2201.08239, 2022.
[12] SUN Y, WANG S, FENG S, et al. ERNIE 3.0:large-scale knowledge enhanced pre-training for language understanding and generation[PP]. arXiv preprint arXiv:2107.02137, 2021.
[13] 王耀祖,李擎,戴张杰,等.大语言模型研究现状与趋势[J]. 工程科学学报, 2024, 46(8):1411-1425.(WANG Y Z, LI Q, DAI Z J, et al. Current status and trends in large language modeling research[J]. Chinese journal of engineering, 2024,46(8):1411-1425.)
[14] 裴炳森,李欣,蒋章涛,等.基于大语言模型的司法文本摘要生成与评价技术研究[J]. 数据与计算发展前沿(中英文),2024, 6(6):62-73.(PEI B S, LI X, JIANG Z T, et al. Research on the generation and evaluation of judicial text summarization based on large language models[J]. Frontiers of data&computing, 2024, 6(6):62-73.)
[15] CUI J, LI Z, YAN Y, et al. Chatlaw:open-source legal large language model with integrated external knowledge bases[PP]. arXiv preprint arXiv:2306.16092, 2023.
[16] ZHOU Z, SHI J X, SONG P X, et al. LawGPT:a Chinese legal knowledge-enhanced large language model[PP]. arXiv preprint arXiv:2406.04614, 2024.
[17] HUANG Q, TAO M, ZHANG C, et al. Lawyer Llama technical report[PP]. arXiv preprint arXiv:2305.15062, 2023.
[18] 刘泓泽,王耀国,唐圣晟,等.医学大语言模型的应用现状与发展趋势研究[J]. 中国数字医学, 2024, 19(8):1-7, 13.(LIU H Z, WANG Y G, TANG S J, et al. Research on the current application and development trend of medical large language model[J]. China digital medicine, 2024, 19(8):1-7, 13.)
[19] KRALJEVIC Z, SHEK A, BEAN D, et al. MedGPT:medical concept prediction from clinical narratives[J]. arXiv preprint arXiv:2107.03134, 2021.
[20] LI Y, LI Z, ZHANG K, et al. ChatDoctor:a medical chat model fine-tuned on a large language model meta-AI(LLaMA)using medical domain knowledge[J]. Cureus, 2023, 15(6):e40895.
[21] ZHANG H, CHEN J, JIANG F, et al. Huatuogpt, towards taming language model to be a doctor[PP]. arXiv preprint arXiv:2305.15075, 2023.
[22] GitHub. Xunzi-LLM-of-Chinese-classics/XunziALLM[EB/OL]. [2025-09-06]. https://github.com/Xunzi-LLM-of-Chineseclassics/XunziALLM.
[23] IHNAINI B, SUN W, CAI Y, et al. Sentiment analysis of Song Dynasty classical poetry using fine-tuned large language models:a study with LLMs[C] //20247th International conference on artificial intelligence and big data. Chengdu:IEEE, 2024:590-597.
[24] MANAPBAYEVA Z, ZAURBEKOVA G, AYAZBEKOVA K, et al. AI in literary translation:ChatGPT-4 vs. professional human translation of Abai’s poem‘Spring’[J]. Procedia computer science, 2024, 251:526-531.
[25] 赵志枭,胡蝶,刘畅,等.人文社科领域中文通用大模型性能评测[J]. 图书情报工作, 2024, 68(13):132-143.(ZHAO Z X,HU D, LIU C, et al. Performance evaluation of Chinese universal large model in the field of humanities and social sciences[J]. Library and information service, 2024, 68(13):132-143.)
[26] 姜鹏,任龑,朱蓓琳.大语言模型在分类标引工作中的应用探索[J]. 农业图书情报学报, 2024, 36(5):32-42.(JIANG P, REN Y, ZHU B L. Exploration and practice of classification indexing combined with large language models[J]. Journal of library and information science in agriculture, 2024, 36(5):32-42.)
[27] 胡开宝,李翼.机器翻译特征及其与人工翻译关系的研究[J]. 中国翻译, 2016, 37(5):10-14.(HU K B, LI Y. Characteristics of machine translation and its relationship with human translation[J]. Chinese translators journal, 2016, 37(5):10-14.)
[28] JEAN S, LAULY S, FIRAT O, et al. Does neural machine translation benefit from larger context?[PP]. arXiv preprint arXiv:1704.05135, 2017.
[29] WU Y, HU G. Exploring prompt engineering with GPT language models for document-level machine translation:insights and findings[C] //KOEHN P, HADDOW B, KOCMI T, et al.Proceedings of the eighth conference on machine translation.Singapore:Association for Computational Linguistics, 2023:166-169.
[30] GAO R, LIN Y, ZHAO N, et al. Machine translation of Chinese classical poetry:a comparison among ChatGPT, Google Translate, and DeepL Translator[J]. Humanities and social sciences communications, 2024, 11(1):835.
[31] JIN K, ZHAO D, LIU W. Morphological and semantic evaluation of ancient Chinese machine translation[C] //Proceedings of the ancient language processing workshop. Bulgaria:INCOMA,2023:96-102.
[32] WANG D, LIN L, ZHAO Z, et al. EvaHan2023:overview of the first international ancient Chinese translation bakeoff[C] //Proceedings of ALT2023:ancient language translation workshop.Macao:Asia-Pacific Association for Machine Translation, 2023:1-14.
[33] MCMANUS S M, LIU R, LI Y, et al. The ups and downs of training RoBERTa-based models on smaller datasets for translation tasks from classical Chinese into modern standard mandarin and modern English[C] //Proceedings of ALT2023:ancient language translation workshop. Macao:Asia-Pacific Association for Machine Translation, 2023:15-22.
[34] WANG J, ZHANG X, LI J, et al. Pre-trained model in ancientChinese-to-modern-Chinese machine translation[C] //Proceedings of ALT2023:ancient language translation workshop. Macao:Asia-Pacific Association for Machine Translation, 2023:23-28.
[35] JU Z, XIN Y, YE M. Machine translation of classical Chinese based on unigram segmentation transformer framework[J]. Applied and computational engineering, 2024, 37:23-30.
[36] 刘洋,王东波.古籍智能信息处理研究现状[J]. 图书情报工作, 2024, 68(23):120-138.(LIU Y, WANG D B. Current research on intelligent information processing for ancient books[J]. Library and information service, 2024, 68(23):120-138.)
[37] 吴梦成,林立涛,吴娜,等.融合不同语义知识的中国古代典籍机器翻译研究[J]. 情报资料工作, 2024, 45(2):97-104.(WU M C, LIN L T, WU N, et al. Research on machine translation of ancient Chinese classics by integrating different semantic knowledge[J]. Information and documentation services, 2024,45(2):97-104.)
[38] ALOWEDI N, MOHAMMED A A, AL-AHDAL H. Artificial intelligence based Arabic-to-English machine versus human translation of poetry:an analytical study of outcomes[J]. Journal of Namibian studies:history politics culture, 2023, 33:1523-1538.
[39] DUNDER I, SELJAN S, PAVLOVSKI M. Automatic machine translation of poetry and a low-resource language pair[J]. 202043rd international convention on information, communication and electronic technology, 2020:1034-1039.
[40] WANG S, WONG D F, YAO J, et al. What is the best way for ChatGPT to translate poetry?[PP]. arXiv preprint arXiv:2406.03450, 2024.
[41] VMIJUNV. VMIJUNV/chinese-poetry-and-prose[EB/OL]. [2025-09-17]. https://github.com/VMIJUNV/chinese-poetry-andprose.
[42] BAI J, BAI S, CHU Y, et al. Qwen technical report[PP]. arXiv preprint arXiv:2309.16609, 2023.
[43] YANG A, XIAO B, WANG B, et al. Baichuan 2:open large-scale language models[PP]. arXiv preprint arXiv:2309.10305, 2023.
[44] 左亮,赵志枭,王东波.基于大语言模型的《四库全书》自动分类研究[J]. 信息资源管理学报, 2024, 14(5):22-35.(ZUO L,ZHAO Z X, WANG D B. A study on automatic categorization of the Siku Quanshu based on a large language model[J]. Journal of information resources management, 2024, 14(5):22-35.)
[45] HU E J, SHEN Y, WALLIS P, et al. Lora:Low-rank adaptation of large language models[PP]. arXiv preprint arXiv:2106.09685,2021.
[46] CHEN A, LOU L, CHEN K, et al. Benchmarking LLMS for translating classical Chinese poetry:evaluating adequacy,fluency, and elegance[PP]. arXiv preprint arXiv:2408.09945,2024.
[47] PAPINENI K, ROUKOS S, WARD T, et al. Bleu:a method for automatic evaluation of machine translation[C] //ISABELLE P,CHARNIAK E, LIN D. Proceedings of the 40th annual meeting of the Association for Computational Linguistics. Philadelphia:Association for Computational Linguistics, 2002:311-318.
[48] LIN C Y. ROUGE:A package for automatic evaluation of summaries[C] //Text summarization branches out. Barcelona:Association for Computational Linguistics, 2004:74-81.
Options
文章导航

/

〈 〉