Automatic Extraction, Indexing, Retrieval and Visualization of Multimodal Meeting Recordings for Knowledge Management Activities

Our knowledge society is faced with an increasing amount of data produced by a great variety of sources which results in diverse multimodal and multimedia data. Such data are characterized by a rich and complex structure in terms of intra- and inter-relationships. Their broad usage and reuse is limited unless sufficient technology-based methods for semantic annotation, extraction, indexing, retrieval and visualization can be developed. In this book chapter we address this problem domain by starting our investigation on general aspects of multimodal information systems and introducing a conceptual architecture for a generalized view about such systems. Based on this knowledge, we further focus on meeting scenarios as an example for a concrete application domain. This application area became increasingly important and a very active multidisciplinary research field over the last years. An overview about relevant research activities in the context of multimodal information systems is given. In the third part of this chapter we are discussing a concrete solution approach and practical experiences made in the MISTRAL research project.