Library and Information Service

Library and Information Service >

2026 , Vol. 70 >Issue 13: 56 - 67

DOI: https://doi.org/10.13266/j.issn.0252-3116.2026.13.005

RESEARCH PAPERS

Scientific Entity Recognition Methods Based on Large Language Models Fine-Tuning

  • Wang Jiamin ,
  • Zhang Tao ,
  • Fang Zichen
Expand
  • School of Economics and Management, Xidian University, Xi'an 710126
Wang Jiamin, Associate Professor, PhD, Master's Supervisor, E-mail:wangjiamin@xidian.edu.cn; Zhang Tao, Master's Candidate; Fang Zichen, Master's Candidate.

Received date: 2025-04-23

  Revised date: 2025-08-31

  Online published: 2026-06-23

Supported by

This work is supported by the National Natural Science Foundation of China project titled “Research on the Identification and Evolution Path of Interdisciplinary Knowledge Structure Based on Text Semantic Understanding” (Grant No. 72304218), and Annual Projects of the Shaanxi Provincial Social Science Fund titled “Research on the Fine-Grained Knowledge Unit Citation Network in Academic Literature” (Grant No. 2023N003).

Fold

Abstract

[Purpose/Significance] This study explores the application of fine-tuned large language models(LLMs) in scientific entity recognition from academic literature, aiming to improve the recognition performance of specific types of scientific entities. [Method/Process] First, the original evaluation dataset was preprocessed by sentence segmentation and entity filtering. Specific prompt templates were designed to convert the preprocessed data into a dataset compatible with the fine-tuning format of LLMs. Subsequently, the Low-Rank Adaptation(LoRA) strategy was employed to conduct supervised fine-tuning on four LLMs from the DeepSeek and Qwen series. These models were compared with state-of-the-art(SOTA) deep learning models and prompt learning methods based on the DeepSeek-R1-671B model. Performance was comprehensively evaluated using precision, recall, and F1-score, combined with efficiency metrics such as training runtime and floating-point operations. Finally, an in-depth analysis of the models' misidentification types was conducted. [Result/Conclusion] Experimental results demonstrate that the fine-tuned LLMs perform significantly in scientific entity recognition tasks, achieving an F1 score of 81.20%, and effectively improving the accuracy of scientific entity recognition.

Key words： scientific entity; named entity recognition; large language models; LoRA fine-tuning; prompt learning

Cite this article

Wang Jiamin , Zhang Tao , Fang Zichen . Scientific Entity Recognition Methods Based on Large Language Models Fine-Tuning[J]. Library and Information Service, 2026 , 70(13) : 56 -67 . DOI: 10.13266/j.issn.0252-3116.2026.13.005

References

[1] 曾建勋.基于海量数字资源的科研关系网络构建探究[J].情报学报, 2013, 32(9):929-935.(ZENG J X. On construction of scholarly relations network based on massive digital resources[J].Journal of the China Society for Scientific and Technical Information, 2013, 32(9):929-935.)
[2] 胡吉颖,谢靖,钱力,等.基于知识图谱的科技大数据知识发现平台建设[J].数据分析与知识发现, 2019, 3(1):55-62.(HU J Y, XIE J, QIAN L, et al. Constructing big data platform for sci-tech knowledge discovery with knowledge graph[J]. Data analysis and knowledge discovery, 2019, 3(1):55-62.)
[3] 王佳敏,陆伟,程齐凯,等.基于细粒度关键词引用网络的领域知识多维分析[J].情报学报, 2022, 41(7):733-744.(WANG J M, LU W, CHENG Q K, et al. Multi-dimensional analysis of domain knowledge based on a fine-grained keyword citation network[J]. Journal of the China Society for Scientific and Technical Information, 2022, 41(7):733-744.)
[4] 钱佳佳,罗卓然,陆伟.基于问题-方法组合的科技论文新颖性度量与创新类型识别[J].图书情报工作, 2021, 65(14):82-89.(QIAN J J, LUO Z R, LU W. Novelty measurement and innovation type identification of scientific literature based on question-method combination[J]. Library and information service, 2021, 65(14):82-89.)
[5] POLAK M P, MORGAN D. Extracting accurate materials data from research papers with conversational language models and prompt engineering[J]. Nature communications, 2024, 15(1):1569.
[6] DAGDELEN J, DUNN A, LEE S, et al. Structured information extraction from scientific text with large language models[J].Nature communications, 2024, 15(1):1418.
[7] CHENG Q, CHEN L, HU Z, et al. A novel prompting method for few-shot NER via LLMs[J]. Natural language processing journal,2024, 8:100099.
[8] HUANG L, YU W, MA W, et al. A survey on hallucination in large language models:principles, taxonomy, challenges, and open questions[J]. ACM transactions on information systems,2025, 43(2):1-55.
[9] HU E J, SHEN Y, WALLIS P, et al. LoRA:low-rank adaptation of large language models[C]//Proceedings of the 10th international conference on learning representations. Boston:OpenReview, 2022.
[10] PRATAP S, SINGH A, GUPTA R, et al. The fine art of fine-tuning:a structured review of advanced LLM fine-tuning techniques[J].Natural language processing journal, 2025, 11:100144.
[11] HU Y, YU J, WANG K, et al. Entity extraction for research and development project knowledge graph[J]. Procedia computer science, 2024, 242:17-24.
[12] 化柏林.针对中文学术文献的情报方法术语抽取[J].现代图书情报技术, 2013(6):68-75.(HUA B L. Extracting information method term from Chinese academic literature[J].New technology of library and information service, 2013(6):68-75.)
[13] KONDO T, NANBA H, TAKEZAWA T, et al. Technical trend analysis by analyzing research papers'titles[C]//Proceedings of the 4th conference on human language technology:challenges for computer science and linguistics. Berlin:Springer, 2009:512-521.
[14] BIKEL D M, SCHWARTZ R, WEISCHEDEL R M. An algorithm that learns what's in a name[J]. Machine learning,1999, 34(1):211-231.
[15] HEFFERNAN K, TEUFEL S. Identifying problems and solutions in scientific text[J]. Scientometrics, 2018, 116(2):1367-1382.
[16] 化柏林,何鸿魏.融合词间关系与CNN的科学实体学术功能分类研究[J].情报科学, 2023, 41(7):90-99.(HUA B L, HE H W. The academic function classification of scientific entity based on the word relationship and CNN[J]. Information science, 2023,41(7):90-99.)
[17] 程齐凯,李信.面向语义出版的学术文本词汇语义功能自动识别[J].数字图书馆论坛, 2017(8):24-31.(CHENG Q K, LI X. Automatic recognition of term function in academic text for semantic publishing[J]. Digital library forum, 2017(8):24-31.)
[18] 鄂海红,张文静,肖思琪,等.深度学习实体关系抽取研究综述[J].软件学报, 2019, 30(6):1793-1818.(E H H, ZHANG W J, XIAO S Q, et al. Survey of entity relationship extraction based on deep learning[J]. Journal of software, 2019, 30(6):1793-1818.)
[19] 陆伟,李鹏程,张国标,等.学术文本词汇功能识别--基于BERT向量化表示的关键词自动分类研究[J].情报学报,2020, 39(12):1320-1329.(LU W, LI P C, ZHANG G B, et al.Recognition of lexical functions in academic texts:automatic classification of keywords based on BERT vectorization[J].Journal of the China Society for Scientific and Technical Information, 2020, 39(12):1320-1329.)
[20] 张国标,李鹏程,陆伟,等.多特征融合的关键词语义功能识别研究[J].图书情报工作, 2021, 65(9):89-96.(ZHANG G B, LI P C, LU W, et al. Research on keyword semantic function recognition based on multi-feature fusion[J]. Library and information service, 2021, 65(9):89-96.)
[21] 王佳敏,房子辰,窦永香.引用功能感知的知识单元引用网络构建与多维分析[J].图书情报工作, 2024, 68(21):133-144.(WANG J M, FANG Z C, DOU Y X. Construction and multidimensional analysis of a citation function-aware knowledge unit citation network[J]. Library and information service, 2024, 68(21):133-144.)
[22] 章成志,谢雨欣,张恒.学术文献全文内容中的方法实体细粒度抽取及演化分析研究[J].情报学报, 2023, 42(8):952-966.(ZHANG C Z, XIE Y X, ZHANG H. Extraction and evolution analysis of fine-grained method entities from full text of academic articles[J]. Journal of the China Society for Scientific andTechnical Information, 2023, 42(8):952-966.)
[23] ZHANG H, ZHANG C, WANG Y. Revealing the technology development of natural language processing:a scientific entitycentric perspective[J]. Information processing&management,2024, 61(1):103574.
[24] 张颖怡,章成志.基于公式化表达脱敏与边界识别加强的学术论文研究问题与方法识别研究[J].情报学报, 2024, 43(6):712-732.(ZHANG Y Y, ZHANG C Z. Identification of problem and method in scientific papers based on formulaic expression desensitization and enhanced boundary recognition[J]. Journal of the China Society for Scientific and Technical Information, 2024,43(6):712-732.)
[25] WANG J, CHENG Q, LU W, et al. A term function-aware keyword citation network method for science mapping analysis[J]. Information processing&management, 2023, 60(4):103405.
[26] 梁佳,张丽萍,闫盛,等.基于大语言模型的命名实体识别研究进展[J].计算机科学与探索, 2024, 18(10):2594-2615.(LIANG J, ZHANG L P, YAN S, et al. Research progress of named entity recognition based on large language model[J].Journal of frontiers of computer science and technology, 2024,18(10):2594-2615.)
[27] 张逸勤,邓三鸿,王东波.基于生成式大语言模型的非遗文本嵌套命名实体识别研究[J].现代情报, 2025, 45(10):26-38.(ZHANG Y Q, DENG S H, WANG D B. Research on nested named entity recognition of intangible cultural heritage texts based on generative language models[J]. Journal of modern information, 2025, 45(10):26-38.)
[28] LIU Z, LUO J. AdaMoLE:fine-tuning large language models with adaptive mixture of low-rank adaptation experts[EB/OL].[2025-11-11]. https://arxiv.org/abs/2405.00361.
[29] WANG L, CHEN S, JIANG L, et al. Parameter-efficient finetuning in large language models:a survey of methodologies[J].Artificial intelligence review, 2025, 58(8):227.
[30] 吴春志,赵玉龙,刘鑫,等.大语言模型微调方法研究综述[J].中文信息学报, 2025, 39(2):1-26.(WU C Z, ZHAO Y L, LIU X,et al. Fine tuning methods for large language models:a survey[J].Journal of Chinese information processing, 2025, 39(2):1-26.)
[31] WEYSSOW M, ZHOU X, KIM K, et al. Exploring parameterefficient fine-tuning techniques for code generation with large language models[J]. ACM transactions on software engineering and methodology, 2025, 34(7):1-25.
[32] 朱珈慧,周潇,王博,等.面向解决"卡脖子"技术困境的关键专利识别研究[J].情报学报, 2025, 44(6):720-735.(ZHU J H, ZHOU X, WANG B, et al. Research on key patent identification for solving the technology dilemma of "neck stuck" [J]. Journal of the China Society for Scientific and Technical Information, 2025, 44(6):720-735.)
[33] 刘畅,张琪,王东波,等.基于大语言模型技术的古籍限定域关系抽取及应用研究[J].情报学报, 2025, 44(2):200-219.(LIU C, ZHANG Q, WANG D B, et al. Research on the extraction and application of ancient books'restricted domain relation based on large language model technology[J]. Journal of the China Society for Scientific and Technical Information, 2025,44(2):200-219.)
[34] 张云秋,殷策.基于大模型的中文电子病历实体自动识别研究[J].数据分析与知识发现,2025, 9(8):100-110.(ZHANG Y Q, YIN C. Automatic recognition of Chinese electronic medical record entities based on LLMs[J]. Data analysis and knowledge discovery, 2025, 9(8):100-110.)
[35] 夏天,于凯,余芊蓉,等.基于生成式大语言模型的古文自动断句与标点研究[J].图书情报工作, 2025, 69(5):59-70.(XIA T, YU K, YU Q R, et al. Automatic sentence segmentation and automatic punctuation for ancient Chinese texts based on generative large language model[J]. Library and information service, 2025, 69(5):59-70.)
[36] 黄永文,马玮璐,鲜国建,等.大语言模型驱动的科学数据自动分类研究[J].情报理论与实践, 2025, 48(6):170-177,194.(HUANG Y W, MA W L, XIAN G J, et al. Large language model-driven automated classification of scientific data[J].Information studies:theory&application. 2025, 48(6):170-177,194.)
[37] LYU K, ZHAO H, GU X, et al. Keeping LLMs aligned after finetuning:the crucial role of prompt templates[C]//Proceedings of the 38th conference on neural information processing systems.New York:Curran Associates Inc, 2024.
[38] 余池,陈亮,许海云,等.基于大语言模型的专利命名实体识别方法研究[J].数据分析与知识发现, 2025, 9(6):47-62.(YU C, CHEN L, XU H Y, et al. Research on patent named entity recognition methods based on large language models[J]. Data analysis and knowledge discovery, 2025, 9(6):47-62.)
[39] GUO D, YANG D, ZHANG H, et al. DeepSeek-R1:incentivizing reasoning capability in LLMs via reinforcement learning[EB/OL].[2025-11-11]. https://arxiv.org/abs/2501.12948.
Options
Outlines

/

〈 〉

    Copyrights © LIS Press Co., Ltd. Reproduction is prohibited without permission.