Cross-Domain Recommendation Approach Based on Topic Modeling and Ontology

Single-domain recommendation systems lack in cross-diversity of items. For a sparse user, it is very difficult to recommend item. But we can utilize the knowledge of other domains to recommend an item to that sparse (unknown user preferences) user. This can be possible with the help of cross-domain recommendation systems. After watching movies, listening to songs or reading books, user reviews them with a specific rating. This data which is generated by a user can be used for recommendation purpose. By using topic modeling techniques like Latent Dirichlet Allocation (LDA) and ontology methods, genres can be extracted from reviews. By passing user reviews to the model, a pattern of words can be obtained which further be mapped to genres by implementing ontological profile approach. Dictionary for the genres can be obtained by crawling similar type of words related to a genre. Book dataset and movie lens dataset are used for our implementation purpose. Evaluated result gives better mean precision in comparison with existing approach based on semantic clustering.