Design And Implementation Of Collaborator Recommendation System Based On Knowledge Graph

Academic communication and collaborations have become more frequent with the development of the computer and information technology that doesn't constrained by geographical locations.Analyses on academic collaborations between scholars has got attentions than before.How to construct the network of scholars effectively utilizing the overloaded information of academic and recommend scholars the proper collaborators to each other,this article proposes a new solution to solve these problems based on the network.An academic knowledge graph-based collaborator recommendation algorithm is proposed in this article which combines the content-based method to recommend partners for users and generates the explanations for recommendations.We extract the information on scholars and papers first to construct the knowledge graph of computer science by the top-down method using Neo4 j database.Then we filter a certain number of scholars for users by calculating the similarity through the network structure on a series of meta paths we defined,who didn't build collaborations before but have high similarity to users in reality,and rank them by calculating their similarity with the scholars who have collaborations with users in early stage.We were inspired by the recommendation algorithm model named item-based and after those two steps we rank the scholars to be recommended from their score of similarity and a list and the explanations of the recommendations are generated by finding the shortest path between two scholars in academic knowledge graph.Our algorithm has been deployed on Web of Scholar as the recommendation module,users can get personalized recommendation after completing their information,the system was designed and implemented for displaying scholars' information which provide reference for researchers.