Logical foundations of knowledge-based recommender systems: a unifying spectrum of alternatives

Abstract The aim of this paper is to provide logical foundations for knowledge-based recommender systems, for which, unlike other problem solving tasks, a comprehensive formalization is not yet available. This goal is justified by the need to compare recommenders based on the way they use knowledge to generate recommendations and, consequently, on the underlying semantics of the recommendation process itself. Moreover, since the here adopted logical formalization has been borrowed from other tasks such as diagnosis, many interesting results and opportunities can be transposed from such tasks to recommendation. While we do not aim at proposing a new recommendation generation technique, we believe that our formalization will be the basis for unifying different approaches to knowledge-based recommendation, revealing their semantics and offering a conceptual framework to compare them. In fact, the framework covers different variations of knowledge-based recommendation, such as context-aware, constraints-based, package and group recommendation, as well as recommendation based on negative preferences.