Share
Many organisations process large quantities of (multimedia) content, with digital information often stored in asset, content or document management systems. A good deal of this information is unstructured, which is a hindrance when searching through this information.
|
TNO has accumulated substantial knowledge on searching through unstructured content, not simply textual but also audio and video content. This is the context for the development of the Novalist Mediamining Toolbox that is able to support existing applications in 'tracking down' the content. The toolbox comprises a set of 'mediamining' functions that can be coupled to existing asset, content or document management systems, either separately or in combination.
For multimedia content use is made of speech recognition or optical character recognition. The toolbox also comprises a number of standard functions for metadata generation:
Automatic rather than manual generation of metadata provides enormous cost savings, especially for multimedia content, which is very labour-intensive and not always a straightforward task for people to accomplish. For instance, differences in background, experience and/or education can result in interpretative differences, which may 'pollute' the metadata. The toolbox has been designed to be coupled to the existing situation and so its impact on the content/asset management system, for instance, is minimal.
An example of the implementation of the toolbox is the Radio 1 news. This demonstration shows how speech recognition can be used on the Radio 1 news journals. This transcription can be searched and a link made to the right time position in the source. Of course, this method can also be used for video content.
