Sentiment Analysis Imported Recommendation Model

Collaborative filtering based recommendation model is classic and efficient,but also exists some problems like user cold start.Utilizing the user preference can help to alleviate user cold start problem of collaborative filtering and improve the accuracy of recommendation.Much information about user preference is hidden in the user reviews,which can be extracted by applying sentiment analysis technology.Therefore,this thesis mainly studies a sentiment analysis imported recommendation model that uses sentiment analysis technology to extract the aspect and user sentiment from user reviews.To further build the user preference model,then combine the user preference and the collaborative filtering to improve the accuracy of recommendation.Main researches in this thesis include:(1)Researches on the aspect-level sentiment analysis approaches.Improve double propagation algorithm and design an improved double propagation based aspect-level sentiment analysis model named IDP-SA to extract product aspects and user sentiment towards these aspects from user reviews as user preference information.(2)Researches on the methods of designing user preference model use user preference information extracted from user reviews.Design an IDP-S A based user preference model SA-UP.Combing SA-UP and collaborative filtering to design a recommendation model SA-UPCF.(3)Design and implementation of a prototype recommendation system.Main contributions of this thesis is proposing a sentiment analysis imported user preference model SA-UP,combing user-aspect preference and product-aspect quality to calculate user-product preference.