文摘
Bibliographic coupling (BC) is an effective measure to estimate the similarity between two scholarly articles (i.e., inter-article similarity between the two articles). It works on out-link references of articles (i.e., those references cited by the articles), and is essential for relatedness analysis and topic clustering of scholarly articles. In this paper, we present a new BC measure DescriptiveBC, which employs the titles of the out-link references to improve BC in two ways: given a target article a, DescriptiveBC provides more accurate information about how (based on numerical inter-article similarity) and why (based on textual descriptive terms) a scholarly article is related to a. Visualization of the information can support the identification, clustering, mapping, and navigation of the related evidence in scientific literature. Empirical evaluation justifies the contributions of DescriptiveBC. Release of the reference titles in each article is thus helpful for the dissemination of research findings in scientific literature, and DescriptiveBC can be incorporated into search engines of scholarly articles to help prospective researchers to navigate through the space of related articles online.