AllMusic’s Tone Intersections

In a pre­vi­ous post about A Flat Hier­ar­chy for Sub­jec­tive mp3 Tags, I described the ardu­ous and mar­gin­al­ly reward­ing task of tag­ging my entire library with as many ‘tones’ tags as All­Mu­sic was able to pro­vide. With foobar2000 0.9 final now less than a week away, these tags may prove use­ful soon enough. But a few weeks ago, impa­tient and curi­ous, I decid­ed to put them to anoth­er use:

tones intersection chart

By cre­at­ing a tones/tones tree struc­ture in foo­bar, I was able to count how often each ‘tone’ inter­sects with every oth­er ‘tone.’ What you see above is the begin­ning of that data col­lec­tion, which I ulti­mate­ly planned to ana­lyze in…some way.

After Googling around for ideas on tag clus­ter­ing, I came across gCLU­TO, a free piece of soft­ware that would, mirac­u­lous­ly, do exact­ly what I need­ed — name­ly, mag­i­cal­ly fig­ure out how best to clus­ter each tag with relat­ed tags. I fig­ured four clus­ters would be a com­fort­able num­ber, based on ear­li­er read­ing I had done on a two-axis the­o­ry of musi­cal emo­tion (intense/relaxed and positive/negative).

topographical cluster visualization

Unfor­tu­nate­ly, my com­put­er sim­ply could­n’t han­dle even con­struct­ing and decon­struct­ing the foo­bar tree with­out freez­ing up for about 45 min­utes each time. Plus, col­lect­ing all this data would have meant hours and hours of work, for a goal whose ben­e­fits weren’t very clear to me at all, as well as a halt in incor­po­rat­ing new down­loads into my library. It was a pret­ty excit­ing cou­ple days while it last­ed though.