The 3C architecture: an XML topic maps-based framework for integrating content, context and common knowledge about multimedia

We present an XML topic maps (XTM) based framework for the integration of three "Cs" of the media: content, context and common (background/domain) knowledge. The framework incorporates content analysis modules such as object detectors, context analysis modules such as named entity extractors and common knowledge analysis modules to extract key pieces of knowledge about media elements such as images, sound recordings, videos, graphic illustrations and animations. The extracted pieces of knowledge are integrated via a uniform and extensible knowledge representation method: XML topic maps. The framework allows for the automatic generation of multimedia queries and the automatic population of multimedia databases. The framework also facilitates conceptual indexing of multimedia elements in digital libraries. Key components of the framework are implemented in an automatic presentation generation system.