Research Of Paper Recommendation Algorithm Based On The Knowledge Difference In Uncertain Graph

In the process of academic research,finding and reading literature is the general work of researchers,how to find valuable papers in a large number of papers become an important problem to be improved.In order to solve this problem,the recommendation system of academic papers has gradually become an emerging application direction of recommender systems.Academic paper is a kind of long text with complex information.Users can acquire knowledge or research achievements in a research field by reading papers.The previous academic papers recommender system often uses LDA or CTM model to analyze the subject,and mining user potential research needs through the analysis of the user's read or published literature list,then recommend corresponding papers to the user.Although it can greatly reduce the difficulty of the searching papers.However,the researchers read literature not only in order to consolidate the original knowledge,more is to learn new knowledge,therefore,if the recommend papers can not only be accepted by user but also contain more new knowledge will bring a better user experience.To improve the user acceptance of new knowledge,to improve the diversity of knowledge and serendipity is the main target of this paper.Based on the theory of knowledge learning and topic model,combined with the characteristics of academic papers recommended for academic studies this paper have done the following work.Firstly,this paper try to analyze the subject association between each subject in the academic research,to reduce the influence of the distribution of knowledge in corpus and to describe the relation between knowledge more clearly,this paper adding a probability model in knowledge map which named uncertain knowledge map.Then formed user knowledge model according to the user's reading documents list and research target description,which contains the user's background knowledge and research target knowledge.And use the knowledge differences between user's background knowledge and research target,proposed learning path search algorithm.Then score papers according to the knowledge importance and the similarity to study paths,the user can not find the background knowledge and the research goal of knowledge differences or unable to get special optimal learning path wereprocessed and analyzed.Experimental results show that this paper proposes the method of use semantic model to mining the document's semantic information,use uncertain knowledge graph catch the relation between the knowledge and then recommend papers according to user's background knowledge has a good effect in reducing the time complexity,improving the researchers' acceptance and satisfaction,meanwhile,increasing the diversity of recommend papers' knowledge and recommend serendipity.