Research On Resource Recommendation Method In Knowledge Community Based On Semantic Analysis

Knowledge community is an important place for people to acquire and study resources in the Internet age.It attracts more and more people's attention with its openess and rapidness.Especially to researchers,knowledge community has become the main way of communication.With the exchange of knowledge and the communication among users,the resources in the knowledge community is growing in geometric number,and it's more and more difficult for people to find the resources which they are interested in.As a result,resource recommendation has become the focus of scholars.Most of current recommendation methods give users recommendations only from the literal matching or frequency angle simply and do not fully take the internal semantic information of resources into account.Semantic analysis method can fully tap the potential of the resources,so as to reflect the content and meaning of resources fully.Therefore,it is one of the problems to be considered in the research of the knowledge community resource sharing that how to give users a recommendation combining with the semantic information of resources.This paper attempts to give recommendations by two semantic analysis methods,LDA theme model and user classification ontology,aiming to improve the evaluation index from various angles.The main content of this article is divided into the following two parts.The first part concerns about the recommendation method of the text resources in the knowledge community based on ontology and LDA theme model.In LDA topic model,text resources are expressed as a kind of probability distribution and word distribution of the corresponding subjects.Therefore,compared with the recommendation based on TF-IDF word frequency statistics,the recommendation based on LDA topic model can get the semantic information of text resources and solve the polysemy and synonym problems effectively.However,the LDA model assumes that theme is independent and does not affect each other.As a result,the recommended results based on LDA topic model are always confined to the same topic,and the recommended serendipity is limited to a certain extent.To solve this problem,ontology is introduced into text resource recommendation based on LDA topic model.Firstly,related topics are achieved according to extended topics of LDA topic model.Then,it considers the distribution probability of the related topics in the text.Finally,text recommendation is conducted combined with weighted related topics.The experimental results on the data set provided by the Cite ULike site show the recommendation of the LDA theme model has been improved significantly after the introduction of the ontology.In the second part,this paper gives a research on the recommendation method of collaborative filtering knowledge community based on user classification ontology.At present,the most widely used method of personalized recommendation is the collaborative filtering method,and the most important step of this method is the similarity calculation.With the number of users increasing,the efficiency of this algorithm becomes very low.In order to solve this problem,this paper proposes a method of collaborative filtering based on user classification.Firstly,the huge users are classified into several groups according to a rule-based classification method.Then,with the guarantee of recommendation accuracy,the local neighbor users are discovered for users.Finally,based on the discovered local neighbors,personalized recommendation is conducted.The experimental verification is carried out on the Movie Lens data set,and the accuracy of user classification and recommendation accuracy is evaluated by F1 and the average absolute error of two indicators.Experimental results show that the collaborative filtering recommendation of user classification ontology reduces the computation quantity of neighbor user identification,and improves the efficiency of the algorithm.