基于作者主题模型的学者聚类与学术影响力评价方法研究

Research on Scholar Clustering and Academic Influence Evaluation Method Based on Author Topic Model

[目的/意义]准确地把握学者的领域研究主题,对学者在该领域的学术影响力进行探测,是开展学者评价工作的一个重要方面。[方法/过程]采用作者主题模型算法,对作者研究内容的主题抽取与表示,构建了作者-主题分布矩阵,并进行作者研究主题聚类,在此基础上分别对作者的主题内合作网络和跨主题合作网络的点度中心度和中介中心度指标进行了测度,以此来反映作者在研究领域的研究专业性和知识创新性,并以图书情报与档案学领域学术期刊2012—2016年发表的45494篇学术论文及其所涉及的35586位作者为实验样本进行了实证研究。[结果/结论]结果表明基于主题内合作网络的点度中心度和跨主题合作网络的中介中心度指标的学者学术影响力评价方法是对经典的普赖斯定律中基于作者发文量的核心作者遴选方法的有效补充。

[Purpose/significance]It is an important aspect of carrying out the evaluation of scholars to accurately grasp the scholars’research topic and compute their academic influence in this field.[Method/process]Used the author topic model to extract the topic of the author’s research content,and constructed the author-topic distribution matrix,and then researched on the scholars clustering.Based on the scholars clustering,the author’s intra-topic cooperation network and cross-topic cooperation network are measured respectively,so as to reflect the author’s research expertise and knowledge innovation in the research field.An empirical study was conducted on the authors of the papers published in academic journals in the field of library information and archives from 2012 to 2016.[Result/conclusion]The result shows that the academic influence evaluation method based on the degree centrality of intra-topic cooperative network and the betweenness centrality of cross-topic cooperative network is an effective supplement to the core author selection method.