Uniform access to multimedia information bases through dynamic taxonomies

The problem of browsing and retrieval in multimedia information bases has been traditionally addressed by two different approaches: the first one uses normal query methods (db or IR queries) on metadata or on a textual description of each item. The second one works on primitive multimedia features (such as color, texture, etc.) and usually tries to find items that are similar to a specific selected item. Neither of these approaches is satisfactory for what is claimed to be the usual end-user task: the exploration of an information base in order to find interesting items. This paper shows how dynamic taxonomies, a model proposed for the intelligent exploration of heterogeneous information bases, can be extended in order to provide guided browsing and retrieval for multimedia information bases and seamlessly integrate access through metadata with methods based on primitive multimedia features. It provides an example of interaction and discusses current research issues that address probabilistic membership and granularity of descriptions.