Multi-aspect Recommendation Method Research Based On Contextual Analysis

With the rapid development and wide application of Internet, the information that user posting was growth explosively, it is became popular that analysis users’ behavior and predict trend through big data. And at this point, personalized recommendation system appears, by recommended product or services that user may interested initiatively, it greatly reduces the difficulty of user found the interested content in mass information himself.Traditional recommendation system is generally based on the analysis of users’ behavior, preferences, personal characteristics, and through a variety of recommendation algorithm, which produce recommendations for the user, and usually assume that the user has been always using the same standard, without considering the influence of the context. But in fact, user trend to change the fond of the same products while in a different context. However, most context-aware recommendation techniques mainly aim at exploiting item-level contextual preferences, and few works attempt to detect the aspect-level contextual preferences. Therefore, in this article, we propose an aspect-level recommendation based on user-generated reviews, by using different contextual weighting strategies, research user’s preferences to products in different context. First, we should analysis the user-generated reviews, extract the aspect that user mentioned, preference condition and context, then build the contextual opinion tuples. And research user’s context-dependent preferences and context-independent preferences, at last, we propose a formula combined those two kinds of user preferences to calculate a score which user give it to a product, and generated the recommendation list, recommended the top-N products in the list to the user. We adopt a real hotel reviews data set in our experiment part, the result shows that our method is capable of capturing users’ contextual preferences and achieving better recommendation accuracy in a certain degree.