I’ll admit a strange fascination with the “recommendation engines” scattered around the on-line world, that take lists of things you like (through purchase or claimed preference), and suggest things you’d probably also like. The bulk of these systems seem to use collaborative filtering software to match the patterns of your choices with a database of other user preferences (ie, Amazon’s “users who bought this book also bought…”).
Pandora Internet Radio uses an altogether different approach…matching artists and songs based on their ‘genetics,’ or the fundamental qualities of the musical performance and composition. Say the folks at Pandora:
Together we set out to capture the essence of music at the most fundamental level. We ended up assembling literally hundreds of musical attributes or “genes” into a very large Music Genome. Taken together these genes capture the unique and magical musical identity of a song — everything from melody, harmony and rhythm, to instrumentation, orchestration, arrangement, lyrics, and of course the rich world of singing and vocal harmony. It’s not about what a band looks like, or what genre they supposedly belong to, or about who buys their records — it’s about what each individual song sounds like.
Each user may not agree with that those fundamental genes are, or how they should be combined. But the discovery process the system allows is rather fun to listen to, and fascinating to explore. Give a listen.
I am a fan of Pandora, but what I find very interesting is how often I don’t like the songs they choose, even though I can hear why they chose them. They will have similar characteristics to the songs I do like (and have indicated I like). They will be by similarly genred artist. Sometimes I even think to myself “wow, I really should like this”…yet I don’t. There is still some unquantifiable quality to my aesthetic that pushes me in one direction but not another, even if the directions are very similar.
You can compare this with last.fm (http://www.last.fm/). Its a little program that continuely listens to what you listen to on your iTune, and recommends songs to you based on other users whose iTune library profile is similar to yours. The difference is where Pandora suggests music to you based on quantified music characterisitcs – music reccomendation by computer via an algorithm, Last.fm is more like Amazon’s “people who bought this also bought…” – music recommendation by people’s actual tastes.
The result is Last.fm sometime pulls suprising rabbits out of the hat, that you may not immediately thing you would like, but you do (or not). There is a lesson here that refers back to previous thread about how arts organisations can effectively utilise their database to increase participation.