It is certainly interesting, but I wonder whether it is possible to extract this kind of metadata from arbitrary multimedia.
Having said that, I do think a subset could be supported. For example, extracting the genre from a video should be possible in the future (I know a colleage that's researching genre detection).
Kind regards,
Raynor