Research On The Methods Of Intelligent Retrieval And Recommendation For Literature

With the rapid growth of the Internet information,Google and Baidu as the the general web search engines,have become an important means for people to obtain information.However,for a large number of literature data on the Internet,research on unified intelligent retrieval for literature is absent.In addition,in terms of the newly constructed retrieval platform for literature,how to effectively solve the “cold start” problem of recommendation system for literature has no great solutions now.For those reasons,research on the methods of literature's intelligent retrieval and recommendation are achieved.The methods and ideas for the construction of the unified intelligent literature's retrieval and recommendation system are provided,which shows the significant theoretic value and application value.Research work mainly includes:(1)The research of methods of the index construction for literature.In order to improve the quality of constructing index,firstly,according to the characteristics of different fields in literature,the fields' configured methods of word segmentation,field indexing and field storage are proposed and secondly,in order to improve the quality of word segmentation and in view of the characteristics of literature's keywords which are mostly professional terms,this paper selects appropriate amount key words and constructs an extend word segmentation dictionary based on the statistical analysis of these key words.In order to improve the efficiency of Lucene index construction,the method of using multi threads to write multi directories is proposed,which solves the problem that the efficiency of indexing is affected by the combination of large capacity index segments in a single directory.(2)The research of methods of the index query for literature.In order to improve the relevance of query and result,different weights are assigned to different fields in the literature and a query expansion method based on Word2 vec is proposed,which can caculate the relevance between the extended phrases and the original query and solve the problem of how to rank multiple extend phrases.In the aspect of index query performance optimization,this paper proposes methods of caching technology and of optimize the query result.(3)The research of methods of literature's recommendation.In order to solve the problem of “cold start” for literature's recommendation system,which lack references,this paper proposes a literature's recommendation method based on frequent item-set of authors.For that literature often has multiple authors,this method can recommend other literature to one user,which are wrote by co-auther of literature attracting this user.Firstly,the ability to handle literature's ID is added to FP-Growth algorithm,so that it could calculate the frequent item-set of authors and corresponding literature's ID.Secondly,on the basis of the candidate literature generated by frequent item-sets,the characteristics of the key words are added to recommendation method.Finally,scores and ranks candidate literature and the Top-N highest score literature are recommended to users.(4)Experimental results and system implementation.The proposed methods are verified and analyzed from the experimental point of view and experiments show that these methods are effective.On the basis of literature's retrieval and recommendation function,the system has also implemented user personal information management and user management,page design and so on.In the end,this paper presents the implementation and deployment of the intelligent retrieval and recommendation system for literature.