A Weighted Method for Citation Network Community Detection

Yunwei Chen,Xue Xiao,Yong Deng,Zhiqiang Zhang

2017-01-01

Abstract:There are many reasons for studying the community structure in citation network, whose underlying community structure may help us understand both the obvious and the more subtle interrelations between subfields, which are so important for understand the research structure and paradigm structure of a subject. However, most methods used for citation network community detection are based on topological structure or content separately. Usually such link-based methods ignore the content of citations, and the content-based methods lack the consideration of citation networks' topological structure, which might be crucial for accurate and meaningful community detection. In this paper, we develop a weighted method to identify the community structure of citation network. The new weighted model incorporates both the topological links and content information in citation networks to find communities. A text similarity method is used for weighting the links in citation networks and then constructs a weighted citation network to carry out community detection analysis. Two case studies had been carried out in two fields of Scientometrics and Evolving Citation Networks. Results revealed that using weighted citation edges with VSM-based nodes similarity to detect communities can improve the modularity Q value, which might mean a better assignment result. Four nodes' movement with the direction of better appropriate of community assignation in the 188 node citation network of the field Scientometrics had also indicates the advantage of our weighted method.