Interactive Exploration of Multi-Dimensional and Hierarchical Information Spaces with Real-Time Preference Elicitation

Current proposals for preference-based information access seem to ignore that users should be acquainted with the information space and the available choices for describing effectively their preferences. Furthermore users rarely formulate complex (preference or plain) queries. The interaction paradigm of Faceted Dynamic Taxonomies (FDT) allows users to explore an information space and to restrict their focus without having to formulate queries. Instead the users can restrict their focus (object set, or set of choices in general) gradually through a simple set of actions, each corresponding to a more refined query (formulated on-the-fly) which can be enacted by a simple click. In this paper we extend this interaction paradigm with actions that allow users to dynamically express their preferences. The proposed model supports progressive preference elicitation, inherited preferences and scope-based resolution of conflicts over single or multi-valued attributes with hierarchically organized values. Finally we elaborate on the algorithmic perspective and the applicability of the model over large information bases.