A Context-Based Audiovisual Representation Model for Audiovisual Information Systems

In this paper we present a contextual representation model of audiovisual (AV) documents for AV information systems. In the rst part, we study AV medium, and show that AV intra-document context is always related to a user task seen as a general description task. We then present the AI-Strata model for AV description: audiovisual units (pieces of AV documents) are annotated with annotation elements described in a knowledge base. The annotation elements are connected at the document level. The whole system being considered as a single graph, we de ne a context of one element as end points of graph-paths starting with this element. In order to control contextual paths, we de ne the notion of potential graphs as graph-patterns instantiated in the general graph. Finally, we show how these graphs are used in the main task of AV information system: navigation, indexing and retrieval.