MAVIS 2: a new approach to content and concept based navigation

A key to many of the problems encountered in information handling is that whenever we are dealing with objects in text, images, videos etc we are dealing with particular representations of those objects or concepts not the concepts themselves. As humans we can quickly get to the concept from the representations, using our prior knowledge of the world, but media processing and indeed artificial intelligence, is not sufficiently well developed for this to be possible automatically in multimedia systems, except in simple restricted domains. In the MAVIS 2 project we are attempting to address these problems. A new architecture for multimedia content and concept based retrieval and navigation is being developed. The architecture not only presents a consistent cross media approach to CBR and CBN but also includes the integration of a multimedia thesaurus which can substantially improve the flexibility and versatility of the multimedia information system. The Multimedia Thesaurus or MMT can be regarded as a traditional digital thesaurus, semantic net or concept graph, showing the relationships between concepts relevant to the application area, and associated with the concepts are representations of the concepts extracted from different media. The MAVIS 2 architecture also involves the integration of intelligent agents. These support navigation by utilising both the media content and semantic concepts in the multimedia thesaurus to develop rapid associations between media based features and the semantic concepts they represent. They are also designed to generate pseudo thesaurus groupings by feature clustering when no semantic relations are available