19^th International CODATA Conference
Category: Data Visualization

Semantic information visualization

Martin Klima (xklima@fel.cvut.cz), Pavel Halabala (halabab@fel.cvut.cz), Pavel Slavik (slavik@fel.cvut.cz)
Czech Technical University in Prague, Czech Republic

In this work we will introduce a system for semantic information visualization and interactive manipulation implemented at CTU Prague. Explicit semantic information is necessary for automatic data processing. Commonly used data formats do not natively include such semantic information. The semantics must be therefore supplied in a separate file. The goal of our system is to enable creation of such semantic metadata for adaptation of multimedia content for use on mobile devices. The adaptation should modify the audio/video and vector graphics data in a logical manner to fulfill the user's requirements while limiting the system resources necessary for data transport and rendering.

The data themselves are designed for representation of the visual aspect only, therefore an additional semantic layer must have been added in a form of external metadata. This is achieved by the introduced system.

The metadata creation itself is an interactive process where the human must be present. Nevertheless the system may support the process by appropriate visualization and assemble tools.

In our case we have concentrated on adaptation of complex 3D (VRML) and 2D (SVG) scenes for browsing and annotating on mobile devices. The named formats need to be adapted with the use of additional semantic metadata in MPEG-7 format. We are introducing a tool for creation of the MPEG-7 description. The system supports direct interconnection between the data (VRML) and the semantic visualization. Thanks to extendable system core supporting a plug-in mechanism we can easily add new visualization components and data formats. The implemented 3D visualization is fully interactive. For more efficient manipulation we have created a set of supportive tools for visualization of semantic relations, relation iteration, filtering and reasoning. The functionality was successfully tested and is currently used in the Mummy Mobile Knowledge Management project (IST-2001-37365).