图书馆杂志

图书馆杂志 ›› 2025, Vol. 44 ›› Issue (412): 98-108.

• 信息管理 • 上一篇    下一篇

基于Sentence-BERT 和时序分析的学者研究主题相似性测度

阮光册 黎心怡 廖紫伊(华东师范大学信息管理系)   

  • 出版日期:2025-08-15 发布日期:2025-09-02
  • 作者简介:

    阮光册 华东师范大学信息管理系,博士,副教授。研究方向:信息分析、文本挖掘。作者贡献:研究选题、研究设计、论文撰写。E-mail cgruan@infor. ecnu. edu. cn上海 200062

    黎心怡 华东师范大学信息管理系,硕士研究生。研究方向:信息分析、文本挖掘。作者贡献:收集数据与实验、撰写论文初稿。 上海 200062

    廖紫伊 华东师范大学信息管理系,硕士研究生。研究方向:信息分析、文本挖掘。作者贡献:收集数据与实验、撰写论文初稿。 上海 200062

Scholar Research Topic Similarity Measurement Based onSentence-BERT and Time Series Analysis

Ruan Guangce Li Xinyi Liao ZiyiDepartment of Information Management East China Normal University   

  • Online:2025-08-15 Published:2025-09-02
  • About author:Ruan Guangce Li Xinyi Liao ZiyiDepartment of Information Management East China Normal University

摘要:

学者研究主题相似性测度,是挖掘潜在合作关系和学术社团发现的基础工作。鉴于学者研究主题动态变化的特点,本文将Sentence-BERT 句向量模型与时序分析相结合,从文本语义层面挖掘学者间研究主题相似性测度,并反映相似性的动态变化特征。实验中,本文采集中国知网( CNKI) 中的文献数据,首先运用Sentence-BERT 模型对文献内容进行向量化处理和相似度计算,随后结合时序变化计算学者间研究主题的时序变化特征,最后通过四象限矩阵图呈现计算结果。通过对比实验,本文方法能够较好地从语义层面识别近期研究主题相似度较高的学者。

关键词: Sentence-BERT, 时间序列, 学者研究主题, 相似性测度

Abstract:

Measuring the similarity of research topics among scholars is fundamental for identifyingpotential collaborations and discovering academic communities. Given the evolving nature of scholarlyresearch interests over time this article integrates the Sentence-BERT sentence vector model withtemporal analysis to assess topic similarity from a semantic perspective and while reflecting temporaldynamics. Literature data were sourced from China National Knowledge Infrastructure CNKI . FirstlySentence-BERT model was used to vectorize and calculate the similarity between research texts. Thenthe temporal changes were incorporated to track the evolution of scholars􀆳 research topics. Finally theresults were presented through a four-quadrant matrix diagram. Through comparative experiments thismethod can effectively identify scholars with high semantic similarity in their recent research.