Most of the multimedia objects distributed over the Internet are still too poorly meta-indexed to be of any use in retrieval tasks. In general these dynamic multimedia objects are manually annotated. The high costs involved in manually indexing multimedia objects, which grow ever more in volume, type and complexity, call for automatic multimedia categorization systems. They should not only be based on natural language processing techniques, but also incorporate content and context based categorization schemata for audio, video and other multimodal media. In order to satisfy all these requirements we propose a web-enabled advanced multimedia system. This system is based on a physical, mathematical and logical framework. We demonstrate how our system automatically generates summarisations of multimedia presentations in an E-learning setting.