| 王佳敏,房子辰,崔蕴学.融合词汇语义功能的学科知识结构识别与演化分析[J].数字图书馆论坛,2026,22(3):20~30 |
| 融合词汇语义功能的学科知识结构识别与演化分析 |
| Integrating Term Function for Identification and Evolutionary Analysis of Discipline Knowledge Structure |
| 投稿时间：2026-01-12 |
| DOI：10.3772/j.issn.1673–2286.2026.03.003 |
| 中文关键词: 词汇语义功能；共现网络；问题–方法；知识结构；演化分析；计算机学科 |
| 英文关键词: Term Function; Co-occurrence Network; Question-Method; Knowledge Structure; Evolution Analysis; Computer Science |
| 基金项目:本研究得到国家自然科学基金青年项目“基于文本语义理解的跨学科知识结构识别与演进路径研究”（编号：72304218）资助。 |
| 作者 | 单位 | | 王佳敏 | 西安电子科技大学经济与管理学院 | | 房子辰 | 西安电子科技大学经济与管理学院 | | 崔蕴学 | 西安电子科技大学经济与管理学院 |
|
| 摘要点击次数: 19 |
| 全文下载次数: 26 |
| 中文摘要: |
| 开展学科知识结构识别和演化分析，有助于把握学科领域发展变化规律，进而推动学科发展。传统基于共词网络的研究通常将关键词视为同质性节点，导致分析粒度较粗、语义层次较低，且难以考察研究问题与研究方法等功能词汇之间的协同演化关系。针对上述局限，本研究通过词汇功能增强知识网络节点的语义信息，以提高知识结构识别和演化分析的精细化程度。首先，基于BERT模型识别学术文献中的研究问题与研究方法，构建融合词汇语义功能的异质性共现网络，并采用社区发现算法识别学科知识结构；其次，从知识结构整体演化与“问题–方法”二元演化两个维度开展主题强度量化分析，并以国际计算机学会数据集为例进行实证研究。实验结果表明，该方法有效识别出计算机学科领域的9个知识社区，系统揭示了各知识社区内研究问题、研究方法及其关联模式；融入二元属性的演化分析进一步捕捉了研究焦点的动态转变及其内在“问题–方法”关联逻辑。本研究提升了知识结构分析的细粒度和可解释程度，为学科知识结构挖掘提供了新的视角和方法。 |
| 英文摘要: |
| Identifying and analyzing the knowledge structure and its evolution in a discipline helps to grasp the patterns of development and change in a field, thereby promoting and guiding its progress. Traditional co-word network-based studies typically treat keywords as homogeneous nodes, leading to coarse analytical granularity and low semantic depth, while also failing to capture the co-evolutionary relationships between functional terms, such as research questions and methods. To address these limitations, this study enhances the semantic information of knowledge network nodes through term function to improve the granularity of knowledge structure identification and evolution analysis. First, the BERT model is employed to identify research questions and methods in academic literature, based on which a heterogeneous co-occurrence network incorporating term functions is constructed, and a community detection algorithm is applied to identify the disciplinary knowledge structure. Second, topic strength quantification is conducted from two dimensions: the overall evolution of the knowledge structure and the binary evolution of question-method pairs, with an empirical study using the ACM (Association for Computing Machinery) dataset. Experimental results show that the proposed method effectively identifies nine knowledge communities in the field of computer science, systematically revealing the research questions, methods, and their association patterns within each community. Furthermore, the evolution analysis incorporating binary attributes captures the dynamic shifts in research foci and the intrinsic question-method association logic. This study enhances the granularity and interpretability of knowledge structure analysis, providing a new perspective and methodology for disciplinary knowledge structure mining. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
