[lusr]

I also came here to suggest a Markov chain generator using your data. I know nothing about music but it seems your analysis also needs to include information about timing (A is followed by B within X ms, C within Y ms, etc.) and some sort of clustering information that would allow choruses to be produced (e.g. identify clusters + cluster timing then information about the chords within the clusters).

Even better would be if you guys released your data and allowed other people to come up with creative ways of analyzing and using it ;) I actually have an idea that I think could work quite well and I'd love to build it but lack the data.