all of the app store, all of the time
Grad school is sometimes about doing ridiculous things. One of the ridiculous things that I’ve done was to download all the text and and all the screenshots from the music section of the iOS app store (as of January), and then classify every one of the 38,750 apps: piano, guitar, dj, bells, zither, and so on.
The results of all this work are in this paper that was been published by Sound & Music Computing 2014.
I’ve released the raw data in advance of the paper on both the IDMIL / McGill website, and here. This includes both the classified and unclassified text data, for those who want to play around with machine classification of text.