Paper Recommendation Based On Scholar Community And Evaluation

Research papers play an important role in technical research and innovation activities. They are used to acquire knowledge, broaden their research areas, and import novel ideas from other areas. However, with the rapid growth of the number of papers, it becomes inconvenient for researchers to search the papers they need, especially for some junior researchers. Junior researchers need classical and authoritative papers of their interesting areas to guide them. The main problem with current academic search engines is that their search results are based on keywords matching which can’t meet individual needs.To solve the problem, this paper uses the concepts and methods of community partition to divide the citation network, and computes the influence of papers in community to assure the quality of recommended papers. Besides, my paper also builds a model to evaluate the quality of recommended results. The details are as follows:1) This paper uses the concepts and methods of community partition and introduces a model to recommend authoritative papers based on the specific community. Above all, this model uses Greedy Clique Expansion Algorithm to discover communities. Then, we study the diffusion of influence based on the specific community. At last, our model uses Paper Rank Algorithm to compute the influence of papers and gets are commendation list. Compared with existing paper recommendation methods, our method narrows the scope of recommended papers, and further improves the recommending speed. Besides, our method improves the quality of recommended papers by ranking papers’ influence.2) To evaluate the recommendation results, this paper introduces an evaluation model of paper recommendation results which use the quality of papers and recommendation matching. For the quality of recommended papers, this article considers SCI factor, cited times, downloaded times, author and funded level, and gives different weights to calculate qualities according to the importance. For the match of recommend the results, this paper builds the users’ interest vector and papers’ feature vector by the tags of communities. Finally, combine the quality of papers and the matching between users and recommended papers to evaluate the recommendation result.