The design of a hybrid semantic recommender system

The majority of existing recommender systems use one or more statistical techniques to recommend content. While such techniques can be very effective, they have a number of restrictions, such as their inability to recommend items based on meaning or relationships between different characteristics of each item. This paper describes the design of a hybrid recommender system that uses a combination of statistical and semantic mechanisms to recommend content. In addition, semantics are used to fine-tune the nature of the recommendation on a context-specific basis. Future extensions based on social networks are also described.