基于超网络模型的研究热点探测与聚类主题描述  

Research on Hot Topics Detection and Clustering Results Description Based on Super Network Model

作　　者：孙海生[1] 

机构地区：[1]聊城大学图书馆,聊城252059

出　　处：《情报杂志》2017年第6期93-98,共6页Journal of Intelligence

摘　　要：[目的/意义]共引分析和共词分析是探测领域研究热点问题常用的文献计量学方法,但是这两种方法都存在一定的局限性。而且,现有研究对共现聚类结果的解读缺乏客观性。针对存在的问题,借鉴超网络理论进行改进研究。[方法/过程]选择最新发表而且最受关注的论文作为样本,提取标识文献内容的特征词,构建超网络模型,根据样本相似度进行聚类分析,计算特征词的描述能力、鉴别能力,识别出聚类主题好的描述符,增强聚类结果解读的客观性。[结果/结论]实证分析结果表明,这种方法能够区分特征词在各个聚类主题中的重要性,有利于提高研究人员对聚类结果解读的客观性和准确性。[ Purpose,/Signifieance] Co-citation analysis and co-word analysis are common bibliometrics methods to detect domain hot topics, but these two methods both have certain limitations. Moreover, interpretations of clustering results lack objectivity in current re- search. In this research, super network theory was used to solve the above problems. [ Method/Process] The latest and the most noticeable papers were selected as samples, and sets of representative terms were extracted. The samples were clustered based on similarity, and de- scriptive power and discriminating power of terms were calculated and used to recognize which terms are the best descriptors of the topics. The objectivity of interpretation of clustering results was improved. [ Result/Conclusion ] The empirical analysis result shows that this method can distinguish important terms to interpret the meaning of clusters; it is helpful to improve the understanding of clustering results more objectively and more accurately.

关 键 词：超网络 文本聚类 描述能力 鉴别能力 

分 类 号：G250[文化科学—图书馆学]