Semantic preference retrieval for querying knowledge bases

This work deals with the problem of automatically creating semantic queries for knowledge bases from preference feedback. Semantic knowledge bases are a good source for retrieving entities for item recommendation. We show that preference decisions are not only based on entities, but also on their corresponding predicate-object relations. By extracting the weights from trained preference models, the weighted predicate-object relations can be stored to a user model. The objective is to use such prototype entities in a general user model to formulate semantic queries for recommendation retrieval.