Research On The Paper Reviewer Recommendation Based On Domain Authority And Activity

With the popularity of Internet technology in various application fields,the amount of information in the network increases exponentially.The traditional method to recommend paper reviewer,which is based on artificial screening,is inefficient.Therefore,using the massive experts'papers data to establish an automated expert recommendation model for paper reviewing and recommend appropriate experts to review the papers,which can improve the efficiency of experts recommendation,and ensure the accuracy and scientific of the experts recommendation.This paper starts from the papers to be reviewed and the evaluation experts,through mining the authority degree and activity degree of the research fields,so that the appropriate experts are recommended to review the paper.The main work of this paper is as follows:(1)Mining expert's authority degree.Firstly,use the number of papers and citation number of papers to calculate the G index of an expert;Secondly,use PageRank algorithm to calculate expert authority degree by establishing expert reference network.Though experiment comparison we found that the calculation result of authority degree is more accurate by using PageRank algorithm.We ranked the results and get the ranking of expert authority degree by using PageRank algorithm.(2)Proposing a forecast method of activity degree with cycle time.Based on the publication time and paper keywords,we construct the keyword co-occurrence matrix with cycle time,and smooth the matrix;Using the gradient descent method to solve the optimal decomposition matrix,normalize the matrix,and finally get the prediction matrix.The values of prediction matrix are used as the activity degree.Through comparative experiments,the results show that this method has better effect.(3)Proposing a method of experts recommendation which combines the authority degree and activity degree of experts.Constructing papers'vector space model of and the evaluation experts' vector space model.By using calculation Method of Similarity to calculate the similarity of the vector,and the ranking of the similarity is obtained by sorting the results.The final ranking results were weighted averaging on the ranking of experts' authority degree and the ranking of similarity.The top(10)experts of the final ranking results were used as the final recommended evaluation experts.Through comparative experiments.We found that the results of the method were more accurate by comparing the results of the other methods.