Research On Academic Literature Recommendation Based On The Multi-dimensions Of Document

In recent years,the explosive growth of academic resources makes it difficult for people to locate the information they need in a large amount of data.Although the keyword search mode can alleviate the problem of information overload to some extent,in the face of massive data,the results of keyword search are still large,and users need to select the documents needed again in the search results.The recommendation system based on user history data can obtain the literature related to user interest more accurately.The existing document recommendation system usually completes recommendation according to the content and domain relevance of the document or collaborative filtering among users,and takes less account of the users' academic level.This makes the recommendation results related to user interest,but some of the results do not meet the academic level and reading needs of users,which leads to low utilization and ineffective recommendation.Therefore,it is of great practical value to study the methods of academic literature recommendation for academic users.The main work and achievements of this dissertation are as follows:(1)Through web crawler technology,online academic literature(2000-2020 AAAI conference papers and Baidu Library Literature)is captured as the data set of this study.(2)The concept dependency tree is constructed according to the entries of Wikipedia classification catalog,and the tree depth of the word in the concept dependency tree is regarded as the concept abstraction level of the word.In the process of feature extraction of academic literature,the structure and content features of academic literature are considered,and the traditional method of calculating vocabulary weight is improved.The TF-IDF algorithm combines the abstract level of vocabulary concept with the paragraph weight of vocabulary.This method can extract the characteristic keywords of academic literature and calculate their weights more reasonably,which is conducive to the calculation of semantic similarity and domain similarity between academic literature.(3)In order to improve the effectiveness of recommendation results,this paper proposes an academic literature recommendation method based on the multi-dimensional of document.According to the differences of users' academic level,users are divided into three levels:primary user,intermediate user and advanced user.Measure document content from seven document dimensions,and the different requirements of different levels of users for literature are mapped to the acceptable range of users for each dimension of the document,So that the literature pushed to users not only meets the user's interest preferences,but also corresponds to the user's academic level.This paper focuses on the recommendation methods of three document dimensions,including:Aiming at the level of document,a recommendation method based on the abstract level of document concept is proposed;Aiming at the difficulty of understanding document content,a recommendation method based on document difficulty is proposed;Aiming at the increasing learning difficulty of users,a recommendation method based on the difficulty span of document learning is proposed.This method can push the academic literature to the users which accords with the user's interest preference and academic level.(4)In order to verify the effectiveness and practicability of the academic literature recommendation method proposed in this dissertation,this dissertation achieves a multi-dimensional document based academic literature recommendation system,and tests the academic literature recommendation method proposed in this dissertation by using the constructed paper data set.The experimental results show that the method of recommending academic literature based on document multi-dimensional has a excellent recommendation effect.