Collaborative Fitering Algorithm Incorporating User’s Reviews And Context Information

The rapid development of the Internet has brought about information overload.Recommender system as a means of information filtering provides an effective way tosolve the problem. With the popularity of Web2.0technologies, more and more sitessupport users who can rate and comment on various contents of products. When userspurchase the products, first of all, they often read a large number of user reviews, and thentend to buy them with higher user evaluation. However, the traditional recommendationalgorithms, unable to premeditate the influence of user reviews and recommendingdifferent projects depending on the user environment, causes unsatisfied withrecommendation results. How to improve the recommendation quality of therecommender system have become the main problem people pay close attention. Thispaper has further conducted deep research of collaborative filtration recommendationalgorithm incorporating user review and context information.First of all, aiming at the problem that the traditional collaborative filtering algorithmcannot premeditate the user reviews, a mining algorithm of user reviews based on thefeature-opinions was proposed. Then we integrate it into the recommendation system aftergaining the user’s favorite level about the products. The algorithm gets the collection offeature views by the syntactic relationship between association rules and statements, andthen acquires the polarity of each feature views combining the sentiwordnet. At last, wecalculate the overall polarity and intensity of the review according to the set polarity ofeach feature views.Secondly, aiming at the problem that the traditional collaborative filtering algorithmcan not recommend by the contextual information of users, a collaborative filteringrecommendation algorithm incorporating the user’s reviews and contextual information isproposed. First, the contextual similarity is calculated based on different types ofcontextual information; then the user reviews polarity value as an evaluation criterionparticipates in the rating similarity calculation; finally the contextual similarity determinesthe proportion of similar user ratings in the prediction rating.Finally, we have compared the performance between the proposed algorithms and other algorithms, and the validity of the proposed algorithm is confirmed. Therecommendation accuracy and precision have a larger increase.