Research On Citation Recommendation System Based On Improved PageRank Algorithm

Nowadays,the number of literature resources is increasing fast,in this situation,there are a lot of scientific and technological documents for reference in scientific research activities.However,the huge number of literature resources increases the difficulty of literature research.Researchers need to spend more time and energy to study and judge the relevance and value of literature,and the efficiency of scientific research is also affected.This phenomenon is called ‘overload of scientific information'.In this context,the relevant research on Citation recommendation emerged.The citation recommendation system can recommend the relevant research results for some subject words or context paragraphs entered by users.At present,the academic community is committed to innovation of citation recommendation methods combining various theories and technologies in order to provide more accurate list of citation recommendations,but few researches on the optimization of the list of citation recommendation from the importance of literature in the citation network.In addition,the existing citation recommendation system or model research,its recommendation basis is mostly the metadata information of literature,which may lead to the omission of incomplete metadata information coverage.In view of the above situation,this paper studies the citation recommendation system based on the improved Page Rank algorithm,which is divided into three parts:(1)Based on the Academic Credit Evaluation theory,the Page Rank algorithm is improved,and ACPage Rank algorithm is proposed,which is used to optimize the ranking of the list of citation recommendations.The core idea of this algorithm is to identify the expert group in the citation network,and adjust the weight of some documents on the citation network according to the citation situation of the expert group.(2)Feature vectors are created using the full text of the document to reduce the likelihood that users will miss citations.Combining ACPage Rank algorithm with LSI model,a citation recommendation system is designed.The system can not only display the title of the list of recommended documents,but also show the most relevant paragraphs in the literature and the author and organization information which studies the most topic.(3)The data set of sample is collected,the system developed in this study is tested,and the experimental results are analyzed and summarized.The experimental results show that LSI model is more suitable for citation recommendation system than TF-IDF model.It can not only solve synonym problem better,but also improve recall rate and precision rate by more than 4% compared with TF-IDF model.The LSI model based on ACPage Rank algorithm can further optimize the effect of literature recommendation,and give priority to the users of high-quality literature cited and approved by the expert group.The recommendation results of the system comprehensively consider the cited situation of the literature,the academic credit of the expert group and other factors,which better reflect the process of literature discovery from the perspective of expert practice,so that the scholars who have just started in a field can also enhance their understanding and understanding of the professional domain with the help of the system.In general,the system can reduce the time users read redundant documents,and improve the efficiency of literature research.