Research On Recommendation Algorithm Based On User Sentiment Analysis And Trust Relationship

Currently,recommendation technology has been widely used in e-commerce and other fields to solve the "information overload" problem.At the same time,the emotional information in user reviews is one of the key factors for further mining user preferences and improving recommendation quality.At present,the sentiment quantification method commonly used in the recommendation field ignores the impact of new words on word segmentation,which leads to inaccurate sentiment value quantification,and it is difficult to establish an sentiment model in a sparse data environment,which seriously affects the quality of recommendation results.This paper combines the user's comment information,social information and other multi-source data to construct a recommendation model based on user sentiment analysis and trust relationship,and then improve the recommendation quality of the recommendation system.The main work and contributions are as follows:(1)In order to solve the problem that the emotional information in user comments is difficult to quantify,a sentiment analysis algorithm based on user comments is proposed.The most critical step of a recommendation algorithm based on user sentiment analysis is the analysis and quantification of sentiment value.First extracts high-interest features based on user comments to builds a new sentiment dictionary,and then uses the postive and negative words to establish a quantitative model of comment sentiment.The experimental results show that,compared with the accuracy of the emotion classification of the basic emotion dictionary,the accuracy of emotion classification of the newly-built emotion dictionary is significantly improved,indicating that the proposed algorithm can more accurately quantify the user's emotion value.(2)In order to solve the problem that it is difficult to establish a recommendation model after quantifying user sentiment information,a hybrid recommendation model based on user sentiment analysis is proposed.Existing recommendation systems mostly use global features and use users' interest in specific topics to model,and lack consideration of user interests and preferences that can be mined in user emotional information.Therefore,this paper proposes a hybrid recommendation model that combines collaborative filtering,content-based filtering and sentiment analysis of user reviews.Through sentiment analysis technology,it is possible to dig deeper into the user preferences contained in each user comment information.In the specific experimental verification process,sentiment analysis data from Twitter and movie metadata are used to recommend movies.Sentiment analysis technology provides information about the subjective preference of the audience to a certain movie.The weighted score fusion method is used to integrate user sentiment value into the recommendation ranking process to improve the recommendation model,which effectively improves the accuracy of the recommendation.(3)In order to solve the problem that the sparse user data makes it difficult to improve the recommendation quality of recommendation system based on sentiment analysis,this paper proposes a recommendation algorithm integrating user sentiment trust based on the proposed sentiment analysis algorithm.First,build a user trust network based on user comment emotional data,then optimize it with the "six degree separation theory" that meets the constraints of reality,and finally mine trusted users based on the similarity of user emotional trust in the trust network to achieve recommendation.Experiments show that the algorithm proposed in this paper outperforms existing similar algorithm in terms of recommendation error,which verifies the rationality of the proposed algorithm.