A context-aware approach to user profiling with interactive preference learning
Pekka Malo, Pyry Siitari · Aaltodoc (Aalto University) · 2010
This paper proposes a context-aware method for user profiling and content retrieval based on interactive preference learning.The method uses a novel combination of ontology-driven context modeling with multiattribute optimization, which allows the system to learn an implicit value function to represent the user's preference system.Due to the domain knowledge brought by ontologies, the system is able to account for the semantic context of the information retrieval task while constructing the user profile.Additionally, a collaborative version of the algorithm is proposed, which is useful when only little or none preference information is available on the active user.In order to demonstrate the approach, we present a personalized business news reader application.The performance of the system is evaluated using Reuters RCV1 corpus.