It’s like data mining, but the data is your music collection. I have roughly 3000 songs in my collection (legal, believe it or not), and I finally joined the iTunes bandwagon. A lot of my music came from a one-year subscription I had to the eMusic downloading service (which was $10/month for unlimited downloads at the time). I got a bunch of jazz this way, as well as some other random stuff that looked interesting–but I didn’t necessarily listen to all of it very much. Now with iTunes shuffling through my entire music collection, I am stochastically discovering interesting music that I might not have come across otherwise. A few of my favorites from the last few days:
Echo and the Bunnymen – Hide & Seek
The Future Sound of London – Divinity
John Cougar – Thundering Hearts
They Might Be Giants – XTC vs. Adam Ant
Arab Strap – Autumnal
Alphaville – Control
Heaven 17 – Dive
Now all we need is a sophisticated music mining algorithm to be developed to wade through those million song databases. The stochastic method is certainly fun, but it’s not terribly efficient.

3 comments
Comments feed for this article
May 2, 2006 at 8:24 am
Anonymous
How would it be different than Last.fm?
May 2, 2006 at 9:34 am
Jacob
Last.fm is a good tool, but it still can only help you with the music that you actually listen to (and perhaps give you suggestions for other things you may like). But it doesn’t help you find songs in your collection that you haven’t listened to but still might enjoy.
It is pretty close, though. You would just need a way to allow Last.fm access to your personal database to make recommendations.
June 17, 2008 at 2:12 am
Hector Hidalgo
Que interesante lo que planteas.
Voy a ver si construyo algo, recuerdame en 1 año más para ver que alcance a hacer pues trabajo en mi tesis de grado en un tema de este tipo.
(al menos es lo que planteo hacer :P )