Research On Citation Recommendation Of Academic Literature By Fusing Structure Function And Citation Function

As depth applications of Internet of things,cloud computing and "Internet plus",users are enjoying the convenience of Internet technology,also fall into the plight of"information overload".In the process of meeting the needs of information,users find himself lost in the ocean of information.Recommendation system,as a kind of personalized information service form,can be used as an real-time recommendation based on user's interest preference,and has been widely used in the major shopping sites.At the same time,with the increase of investment in construction of digital resources of different organizations,users also encounter some difficulties in enjoying convenience of digital literature resources,such as literature search is time-consuming and the query results are not satisfactory.Therefore,citation recommendation of academic literature has attracted the attention of scholars,this study can improve the efficiency and quality during writing of users in a certain extent,the ultimate purpose is to automatically find the position of user manuscripts that need support by references,and provide relevant documents.Different general recommendation system,service object of citation recommendation is generally researchers,commendation content is academic literature,and recommended process involves many factors of citation analysis,citation motivation,author's own characteristics.Therefore,recommendation system is not only the specific application of recommendation system,but also should be regarded as the extension and expansion of the recommendation system.Research on citation recommendation has been studied for many years abroad,and domestic research has just started,research is mainly done from the improved recommendation algorithm,and little research focuses on citation recommendation from perspective of structure function and citation function.Based on this,this paper takes structure function and citation function as the starting point of the research,and combines the traditional recommendation methods to improve the effect of citation recommendation.This paper is divided into 7 chapters,the main contents are as follows:Chapter 1 is for Introduction.Firstly this chapter explains the background and meaning.And on the basis of extensive literature research,this chapter makes literature reviews for local recommendation and global citation recommendation,summarizes citation recommendation methods,reviews the current research situation of citation recommendation,and explain the feasibility of this research.Then,on the basis of the above work,this chapter puts forward research ideas and methods as well as research content and innovation in this paper.Chapter 2 introduces related theoretical research on citation recommendation.Starting with the introduction of the general recommendation system,this paper defines the citation recommendation system,including the definition of the citation recommendation,the characteristics of the citation recommendation,and the classification of the recommendation.Then,it introduces the process of citation recommendation through four modules,so as to distinguish the working process of the system with general recommendation system.Next,the evaluation of the results of citation recommendation is based on the method of information retrieval and the characteristics of citation recommendation.Because of the particularity of citation recommendation,the paper also analyzes the data sets used in the experiment of citation recommendation,as well as the user behavior which is different from the general recommendation system.Finally,based on the close relationship between citation context and citation recommendation,this paper makes a summary of the research on citation context.Chapter 3 is the theoretical framework of citation recommendation.In this paper,we firstly propose a theoretical framework of citation recommendation.Then the paper focuses on the motivation and overall process of citation recommendation presented in this paper.It mainly includes the citation recommendation based on the structure function and citation function,and fusion recommendation based on the information fusion method.Chapter 4 is the citation recommendation based on structure function of academic literature.Firstly,the structure function of the document are divided,and then the citation context is used as query to match different structure functions,including local weighted matching and global weighted matching.Local weighted matching is matching between the full text and a part of the structure which is determined by the similarity in advance with query,while the global weighted matching is the overall matching between the different structural functions of literature by query.At the same time,the contribution of different structure function to query is analyzed in this paper.Chapter 5 is citation recommendation of academic literature based on Citation function.The main idea is to make it clear that the probability of generating a corresponding query is higher in citation context(equivalent to an abstract representation of a document).The study involves two methods,one is,based on the preliminary recommendation in the citation,representing document using citation context and then recommend results by two times,the other one is expressing the document directly through the citation context,and then recommend results adopting traditional methods of recommendation.The results show that the reorganization of document sets can improve the effect of citation recommendation.Chapter 6 is citation recommendation based on combination of structure function and citation function.Based on the information fusion method of scoring and ranking,this paper mainly studies fusion recommendation of structure function and citation recommendation,which is mainly introduced from three aspects.The first part is that query is located to different domains of structure function,and fusion of each recommendation results is done in different dimensions;The second part is fusion of the recommendation result chapter 5 and chapter 6 in the study;The third partris that pseudo relevance feedback technology is used to more fully train recommendation model,so as to improve the effect of fusion recommendation.Chapter 7 summarizes the whole work of research in this paper.On the basis of a comprehensive summary of this study,this paper puts forward the deficiencies of this study and related follow-up studies.