Ranking Query Results using Context-Aware Preferences

To belter serve users' information needs without requiring comprehensive queries from users, a simple yet effective technique is to explore the preferences of users. Since these preferences can differ for each context of the user, we introduce context-aware preferences. To anchor the semantics of context-aware preferences in a traditional probabilistic model of information retrieval, we present a semantics for context-aware preferences based on the history of the user. An advantage of this approach is that the inherent uncertainty of context information, due to the fact that context information is often acquired through sensors, can be easily integrated in the model. To demonstrate the feasibility of our approach and current bottlenecks we provide a naive implementation of our technique based on database views.