The MPEG-7 standard is currently being developed as a way to unify the description of multimedia asset content to support searching analogous to the way we can currently search text documents. To help guide this ambitious task, the organizing committee has been putting together a series of symposia at which potentially relevant research is described. I was asked to talk at the most recent event (in San Jose) on computational auditory scene analysis, which could be useful for generating automatic content description of audio documents.
You can view the talk slides in Acrobat PDF format. If you use the Acrobat web-browser plug-in, you should be able to click on the spectrogram images to hear the associated sounds. (This won't work if you download the PDF and view it locally, since the sound files are on my web server, not encapsulated within the PDF).