Community and Relation Discovery in E-mail Network Using Author-Recipient-Community Topic (ARTC) Model
2015-07-15 15:05:16   来源：   评论：0 点击：

Abstract—-In recent years, the issue of community detection has come into prominence and many approaches have been developed to extract latent communities by analyzing social documents and connections. Previous works have largely focused on the topology of networks. However, recent studies deal with topic-based community detection. The drawback of topic-based methods is that they obtain topologically diverse sub communities, because they concern with content analysis of nodes and ignore link information of the network. However, edges have information about properties of communities, because they are important points to analyze the content of relationships. In this work, we have proposed a model to learn communities sharing similar topics and to identify community distributions based on the content of edges using messages sent between users. The proposed model is built on Community Topic Model (CTM) by adding the link information of corpus. Experimental studies on Enron e-mail corpus have shown that proposed model was able to group edges in a community in regard to similar topics. The proposed model analyses relations between users in the same community and these communities appear a part of the social network of Enron. Also, top actors have been identified by using ARCT model.

Index Terms—community detection, unsupervised learning, text mining.

Cite: Arzu Kakisim and İbrahim Sogukpinar, "Community and Relation Discovery in E-mail Network Using Author-Recipient-Community Topic (ARTC) Model," Lecture Notes on Information Theory, Vol. 3, No. 1, pp. 1-7, June 2015. doi: 10.18178/lnit.3.1.1-7