Keyword-driven Citation Recommendations

Nowadays,paper recommender system often recommends academic papers based on users'personalized retrieval demands.Typically,the paper recommender system analyzes the keywords typed by a user and then returns his or her preferred papers,in an efficient and economic manner.In practice,one academic paper often contains partial keywords that the user is interested in.Therefore,the paper recommender system needs to return the user a set of papers that collectively covers the users'query keywords.However,the paper recommender system only uses the exact keyword matching technique for recommendation decisions,while neglecting the correlation among papers'research content.As a consequence,it may output a set of papers from multiple disciplines that are different from the user's real research field;furthermore,these output papers fail to satisfy the user potential requirements on deep and continuous research on a certain domain or topic.Fortunately,an existing literature citation graph that depict the citation relationships among different papers have provided a promising way to model the papers'correlations?the correlation relationships?.Therefore,the paper proposes and designs two different paper recommendation approaches based on the existing literature citation graph.The main research contents of these two approaches are as follows:?1?Keywords-driven and popularity-aware paper recommendation approach based on an undirected literature citation graph.Currently,an existing paper recommender system mainly uses the keywords entered by users to recommend papers,but the paper recommendation process ignores the correlation among papers'research content.In view of this shortcoming,the paper firstly proposes a keywords-driven and popularity-aware paper recommendation approach based on an undirected literature citation graph,named PRkeyword+pop?Paper Recommendation?approach.PRkeyword+pop approach can search for a set of papers on the undirected literature citation graph according to the users'query keywords;furthermore,this approach considers not only the correlation relationships among papers but also the papers'popularity in the paper recommendation process.In addition,the paper conducts large-scale experiments on the real-life Hep-Th dataset to further demonstrate the usefulness and feasibility of PRkeyword+pop approach.Experiment results demonstrate that,with other paper recommendation approaches compared,PRkeyword+pop approach can search a set of satisfactory papers for users.?2?Keywords-driven and weight-aware paper recommendation approach based on a weighted paper correlation graph.Although an existing literature citation graph depicts citation relationships among different papers,the citation relationships of the existing literature citation graph are very sparse.Furthermore,the existing literature citation graph does not consider the possible self-citations from papers'authors.Considering the above issues,the paper puts forward a weighted similarity-based link prediction approach,the approach combines paper publication time,paper keywords and paper authors information for optimizing the existing literature citation graph;thus,the paper constructs a weighted paper correlation graph based on the existing literature citation graph.The paper again proposes a new paper recommendation approach based on the weighted paper correlation graph,named LP-PR_{keyword+weight}?Link Prediction-Paper Recommendation?approach.LP-PRkeyword₊weight approach can search for a set of papers on the weighted paper correlation graph according to users'query keywords;furthermore,this approach considers not only the correlation relationships among papers but also the strength of the papers'correlation?i.e.,weight?in the paper recommendation process.Likewise,the paper also conducts large-scale experiments on the real-life Hep-Th dataset to further demonstrate the usefulness and feasibility of LP-PRkeyword₊weight approach.Experimental results prove the advantages of LP-PRkeyword₊weight approach in searching for a set of satisfactory papers compared to other competitive approaches.