Does that work, distinguishing spoken word from music? That would be extremely useful.

I’m not sure about that namespace recommendation though – just makes it harder to process, in my experience.