On using reviews and comments for cross domain recommendations and decision making

To help people make choices and take decisions is an important function of recommender system. Reviews and comments that are written online by users after watching movies, reading books, listening to music or purchasing a specific item, are important sources of user generated data, which can be utilized for decision making using recommendations given to user by the recommender system. When the available data in one domain is sparse, data (content and ratings) from other domains can be used to make cross domain recommendations. In this paper we have used this content information as well as ratings of both domains where there is no user-item overlap between the given domains for cross domain recommendations. User generated content (reviews and comments) crawled from web requires topic modeling to discover the latent thematic structure in the corpora of both domains. Since the topics in both domains are dissimilar, we have compared various approaches based on semantic space of corpus and knowledge based methods for finding cross domain recommendations. Experimental results show improvement in precision in recommendations over existing approach based on semantic clustering.