Semantic-based Multimedia Representations for the Museum Experience

We introduce a common architecture to store semantically enriched multimedia material that can be accessed by intelligent engines able to automatically assemble video clips and build textbased hypermedia. The core of the architecture is the Multimedia Database (MMDB), which stores a repository of multimedia information (i.e. text, images, speech, audio, etc.) and its annotation with respect to a shallow representation of its meaning. We present the architecture of the system, the computational framework for annotation, the applications that can access the data repository and we discuss some preliminary results of the evaluation.