Expert Search Based On Characteristics Of Academic Social Network

At present, the academic development change rapidly, those academic activities and communications occurred frequently, the number of literatures growth rapidly, As for academic search in a large amount of data, searching literatures efficiently and locating experts rapidly have become the urgent problem for academic researchers. The expert search is from searching the user needs of academic experts from large numbers of literatures and scholars such as heterogeneous data in an efficiently and accurately way, academic expert search gradually become a new hot spot for researchers. Existing expert search commonly used is based on the text information retrieval, this retrieval method query related literatures though entering the query words by users. General we can not directly determine whether the author is a field of academic experts, and the definition of experts is different among people, so there are different opinions of experts, people can only rely on the web site or the recommendation of academic circles. The expert search model includes probabilistic topic model, language model and PageRank, these methods are based on the search text or links among literatures. Along with the academic development, the relationship between academic trends to be diversification, there is a big limitation to search an expert only from the text or links angle mode, there must have text data to retrieval the relationship between scholars, while the relation of experts also reveal status and authority of a scholar, we can not only search experts from the perspective of literatures, but also evaluate expert from the perspective.As for the defects of the existing academic expert search algorithms only rely on one source of data to search an expert, Without consideration the others characteristics on the impact of an expert, a method that based on the characteristics of academic social network named ANF was proposed in this paper, which fully considered three heterogeneous data of citation network, co-author network and contents of papers. ANF algorithm measures the value of an expert from the papers’ times cited, co-citation rates and authority values in citation network, and analysis the centrality of an expert from the global and local aspects in co-author network at the same time, combining text mining method to calculate the similarity between the users’ query and papers’ contents. Moreover we use BP neural network to decide the ranking of those experts. The result of experiments shows the ANF verify the effectiveness of search academic expert based on the characteristics of the academic society, then we draw the topological graph and get relevant experts ranking in the field of academic and topic hot words, compared with the existing expert search algorithms and improved the accuracy of expert search.