CICPV: A New Academic Expert Search Model

Academic expert search is one of the most important issues for mining academic networks. There exist the different types of information (e.g. papers, authors and citations) in an academic network. Different from the traditional academic expert search models, this paper makes good use of the important data from the citation network, the coauthor network and the papers' content by introducing social network analysis method. Our model can measure an expert's citation influence using the citation ratio, co-citation ratio and authority value of his papers in the citation network. It analyses an expert's centrality from the global and local aspects of the coauthor network, also combining the text mining method to calculate the similarity between the users' query and papers' contents. In order to accurately express a paper, we improve the VSM model by adding the location weights. Moreover, our model uses the BP neural network to decide the ranking of each expert. Experimental results show that our method can improve the performance of academic expert search.