Study And Implementation Of High Quality Academic Resource Recommendation

With the rapid development of Internet, searching information on Internet has become an important way for people. However, facing the huge number of online academic resources with different quality, how to allow users to get academic resources they interested and intelligently filter out the low quality resources has become a serious problem.This paper mainly studies user interest expression, authoritative value calculation and academic recommendations, in order to recommend more valuable academic resources for users. On the basis of studying the user model and the paper model building method, author and authority values calculation method and academic papers recommend methods, the paper proposes a three layer graph model, an academic authority value algorithm, and the corresponding personalized academic resources recommend scheme. Then this paper implements the academic recommend scheme, determines the parameter values in the scheme through experiments, and analyzes and summarizes the experiment results.The scheme proposed in this thesis includes three-layer graph model, user interest model, paper quality and author authority calculation method and academic recommendation method based on collaborative filtering. Three-layer graph model contains user layer, paper layer, topic layer and the relationship between layers. The model builds the indirect relationship between users and topics through the relationship between users and papers built by users’ operation behaviors on papers and the relationship between papers and topics. User interest model is used to represent users’ preference. The user interest model presented in this paper contains a collection of the user’s current research topic, the user title space feature vector, the user abstract space feature vector and user keyword space feature vector. Author authority value and paper quality value is calculated based on the relationship between authors and their papers published, the author level, the paper citation times, the publication time and the journal/conference level. On the basis of the three-layer graph model, using the user-based collaborative filtering algorithm, we can recommend research field for users. According to the three-layer graph model and paper content, we can build user interest model. Using the author authority value and paper quality value calculation method, we can calculate author authority value and paper quality value. Then through calculating the similarity between users and authore/papers, we can recommend authority authors and high quality papers for users.Experimental results show that the proposed user interest model can better express users’ interest, the author authority and the paper quality calculation method can increase the portion of authoritative authors and high quality papers in recommender, and the research field recommend based on three-layer graph model can dig out the research direction for users.