An Academic Social Network Search Algorithm Based On Correlation Analysis

Academic social network is constructed by academic activities, and the nodes in the network are represented scholars, and co-citation relation among scholars constitutes the network sides. With the development of academic research, the size of academic social is growing. Searching for the information you need in such a vast amount of information is a relatively new research topic. Currently, there are many scholars to study the academic social search, and have yielded progress. Putting such scholar search into practice, paper-reviewer recommendation is a typical kind of application. It considers social relation between reviewers and by-reviewers based on the expert search in order to search out qualified specialists. In order to solve the problem related to academic social network search, this paper proposes a kind of academic social network search algorithm based on correlation analysis.In this paper, the algorithm mainly includes: firstly, content correlation between candidate node and query node needs to be calculated. It refers similarity among short texts.In order to make calculation results of content similarity more comprehensive and more in line with the actual situation, a kind of semantic association calculation model based on neighbor nodes is proposed, thus some weakness can be fixed. Secondly, structure correlation between candidate node and query node nodes to be calculated, and it is represented by shortest path distance among nodes. The network graph is kind of not-right diagram, thus the shortest path between any nodes is calculated via BFS. Thirdly, authority of candidate node needs to be calculated. Comprehending three factors above, correlation model is constructed between candidate node and query node. Finally, random walk method is used to search node.In order to make search progress quicker, random walk method based on shortest path distance is proposed. Therefore, every node has total score. Based on the score, some destination nodes are returned to user.The datasets used in this paper are from C-DBLP. The experiments results show that the search method proposed in this paper outperform other methods, and are consistent with previous theories postulate.