基于语义信息的术语加权算法提升科技文献检索的准确性

图书馆杂志 ›› 2025, Vol. 44 ›› Issue (415): 28-39.

基于语义信息的术语加权算法提升科技文献检索的准确性

张敏1, 4 李唯2 范青3[1 中国科学院武汉文献情报中心 2 武汉软件工程职业学院(武汉开放大学) 3 华中师范大学国家文化产业研究中心 4 科技大数据湖北省重点实验室]

出版日期:2025-11-15 发布日期:2025-11-26
作者简介:
张敏中国科学院武汉文献情报中心,工程师, 博士。研究方向:信息抽取、语义索引、深度学习等。作者贡献:文献调研、设计实验、论文撰写及修改。E-mailzhangmin2012@mail. whlib. ac. cn 湖北武汉430071
李唯武汉软件工程职业学院(武汉开放大学),教授。研究方向:深度学习、数据挖掘与分析。作者贡献:设计研究方案,撰写部分论文,修改和完善论文。湖北武汉430205
范青华中师范大学国家文化产业研究中心,副教授。研究方向: 知识工程、数据挖掘与分析。作者贡献:参与实验工作及评价。湖北武汉 430079

Improving the Accuracy of Scientific Literature Retrieval through Term Weighting Algorithms Based on Semantic Information

Online:2025-11-15 Published:2025-11-26
About author:
Zhang Min1 4 Li Wei2 Fan Qing3（1 Wuhan Library Chinese Academy of Sciences 2 Wuhan Vocational College of Software andEngineering Wuhan Open University 3 National Cultural Industry Research Center of Central ChinaNormal University 4 Hubei Key Laboratory of Big Data in Science and Technology）

摘要/Abstract

摘要：

传统的术语加权算法常常忽略了科技文献中术语的语义信息,从而无法准确评估术语的重要度,导致科技文献检索结果的准确性不高。为了充分利用术语的语义信息,本文提出了一种基于语义信息的术语加权算法,旨在提高科技文献检索的准确性。本文提出的算法将使用语义信息权重来衡量术语的语义信息重要度。同时,该算法还会基于TF-IDF 算法计算术语的关键词权重。通过将这两种权重进行加权,得到一种综合的术语权重来衡量术语的重要度。在实验中,本文提出的术语加权算法相比于传统的TF-IDF 和BM25 算法表现出了更好的效果,能够有效地提高科技文献检索的准确性。

关键词: 术语加权&emsp, 语义信息权重&emsp, 检索准确性&emsp, 科技文献检索&emsp, TF-IDF

Abstract:

Traditional term weighting algorithms often overlook the semantic information of terms inscientific literature thereby failing to accurately assess the importance of terms and leading to lowaccuracy in the retrieval of scientific literature. In order to fully utilize the semantic information of termsthis paper proposes a semantic-based term weighting algorithm aimed at improving the accuracy ofretrieval in scientific literature. The proposed algorithm will utilize semantic information weights tomeasure the importance of term􀆳s semantic information. Simultaneously the algorithm will also calculatethe keyword weight of terms based on the TF-IDF algorithm. By combining these two weights acomprehensive term weight is obtained to gauge the importance of the term. In the experiments the termweighting algorithm proposed in this paper demonstrated better performance compared to traditional TFIDFand BM25 algorithms effectively enhancing the accuracy of retrieval in scientific literature.

Key words:

Term weighting, Semantic information weight, Search accuracy, Scientific literatureretrieval, TF-IDF

张　敏1, 4 　李　唯2 　范　青3[1 中国科学院武汉文献情报中心　2 武汉软件工程职业学院(武汉开放大学) 　3 华中师范大学国家文化产业研究中心　4 科技大数据湖北省重点实验室]. 基于语义信息的术语加权算法提升科技文献检索的准确性[J]. 图书馆杂志, 2025, 44(415): 28-39.

Zhang Min1 4 Li Wei2 Fan Qing3（1 Wuhan Library Chinese Academy of Sciences 2 Wuhan Vocational College of Software andEngineering Wuhan Open University 3 National Cultural Industry Research Center of Central ChinaNormal University 4 Hubei Key Laboratory of Big Data in Science and Technology）. Improving the Accuracy of Scientific Literature Retrieval through Term Weighting Algorithms Based on Semantic Information[J]. Libraly Journal, 2025, 44(415): 28-39.

参考文献

[ 1 ] Yu B. Research on information retrieval modelbased on ontology J . EURASIP Journal onWireless Communications and Networking 20192019 1 1 8.

[ 2 ] Salton G Yang C S Wong A. A vector spacemodel for information retrieval J . Journal of theAmerican Society for Information Science 197518 11 613 620.

[ 3 ] Robertson S E Walker S. Some simple effectiveapproximations to the 2-Poisson model forprobabilistic weighted retrieval J . Proceedings ofthe 17th Annual International ACM SIGIRConference on Research and Development inInformation Retrieval. New York Springer-VerlagNew York Inc. 1994 232 241.

[ 4 ] Li Q Lee S Jung H et al. Term weighting forinformation retrieval based on term??s discriminationpower J . Multimedia Tools and Applications2014 71 2 769 781.

[ 5 ] Zamani S Lee S P Shokripour R et al. A nounbasedapproach to feature location using time-awareterm-weighting J . Information and SoftwareTechnology 2014 56 8 991 1011.

[ 6 ] Ibrahim O Landa-Silva D. Term frequency withaverage term occurrences for textual informationretrieval J . Soft Computing 2016 20 8 30453061. [ 7 ] Paik J H. A novel TF-IDF weighting scheme foreffective ranking C / / Proceedings of the 36thinternational ACM SIGIR conference on Researchand development in information retrieval. 2013 343352. [ 8 ] Boubekeur F Azzoug W. Concept-based indexingin text information retrieval J . arXiv PreprintarXiv 1303. 1703 2013. [ 9 ] Kumari M Jain A Bhatia A. Synonyms basedterm weighting scheme An extension to TF. IDFJ . Procedia Computer Science 2016 89 555561.

[10] Dai Z Callan J. Context-aware document termweighting for ad-hoc search C / / Proceedings ofThe Web Conference 2020. 2020 1897 1907.

[11] Blanco R Lioma C. Graph-based term weightingfor information retrieval J . Information Retrieval2012 15 1 54 92.

[12] Boughanem M Mallak I Prade H. A new factorfor computing the relevance of a document to aquery C / / International Conference on FuzzySystems. IEEE 2010 1 6.

[13] 李祯静,秦春秀,赵捧未,等. 科技文献的资源语义空间:一种细粒度知识组织方法[J]. 情报杂志,2019, 38(2):158 165; 180.

[14] 秦春秀,郑梦悦,马续补,等. 基于智能主题图的科技文献细粒度知识组织模型[J]. 情报杂志, 2021, 40(11):169 175; 135.

[15] 张敏,丁良萍,刘欢. 面向科技文献的多维语义索引构建思路及实现[J]. 情报理论与实践,2021,44(8):139 145.

[16] Swales J M. Research Genres Explorations andApplications M . Cambridge Cambridge UniversityPress 2004 228 229.

[17] 张智雄,刘欢,丁良萍,等. 不同深度学习模型的科技论文摘要语步识别效果对比研究[J]. 数据分析与知识发现,2019, 3(12):1 9.

[18] 王末,崔运鹏,陈丽,等. 基于深度学习的学术论文语步结构分类方法研究[J]. 数据分析与知识发现,2020, 4(6):60 68.

[19] arXiv 官网地址和简介[2023-01-05]. https / /arxiv. org. [20] ScienceWise 官网地址和简介[2024-01-10]. https / /sciencewise. info / .

[21 ] Zeynali Tazehkandi M. Review of ranked-basedand unranked-based metrics for determining theeffectiveness of search engines J . HumanInformation Interaction 2020 7 2 1 15.

[1]	刁羽薛红（四川轻化工大学图书馆）. 基于电子资源行为数据的TF-IDF文献推荐方法研究——以电子资源校外访问系统为例 [J]. 图书馆杂志, 2022, 41(12): 45-54.
[2]	任树怀. LUCENE搜索算法剖析及优化研究[J]. 图书馆杂志, 2014, 33(12): 17-23.