Save the Robot – Chris Dahlen

Music recommendation engines fascinate me. The idea that a piece of software could study what I listen to, do some math, and suggest some other music that I might like, is a perfect example of how cold hard logic can rub against ephemeral, impossible-to-quantify taste. I’ve seen die-hard music fans get incredibly upset at the idea that a computer could understand what they like about music. They tell you that a computer doesn’t know how the blues make you feel, a computer doesn’t care that Bruce Springsteen was on the radio the first time you got laid in a Corvette. They tell you all kinds of things. And I just sit there and smile and think, “Okay – but have you tried it?”

A couple years ago I wrote a lengthy feature on music recommender engines, highlighting some of the techniques and major companies that were exploring the space. Many more people have jumped in, and one of the folks I interviewed – Sun Microsystem’s Paul Lamere – works in this space and chronicles it on his blog. I bring it up to point out something he’s been writing about recently: making your musical preferences portable.

Most of the recommender engines out there – Pandora and Last.fm are still two of the leaders – study you and collect data on you, but keep it for their own purposes. Paul has been researching APML, a markup language that lets you store and share your personal interests. It stands for Attention Profile Markup Language, and as Paul explains it:

Lately there’s been quite a bit of attention being paid to making sure that the data that describes the things that we like, our attention data, is portable. With portable attention data, we could go to any music store and be directed to the music that we are most likely to want to listen to. We won’t have to spend any time rating tracks or artists, we’ll just show the music store our taste data. Of course, this taste data needs to be in some standard format so that everyone can understand it. One effort at standardizing our taste data is APML. APML is an XML based language that allows users to share their own personal taste data in much the same way that OPML allows the exchange of reading lists between blog readers. APML is new and not finished yet, but even in its infant state, it is garnering lots of support.

I am particularly interested in how APML could be used to represent an individual’s music taste. One possibility is to have the APML file for the individual list the artists that a person likes (or vehemently dislikes). Another approach is to have the preferences be more abstract – to list weighted affinities toward music genres or styles. The latter approach seemed much more interesting to me – it offers some bit of privacy (instead of seeing Paris Hilton in my APML file, you would just see Female Pop Singer).

One company’s already giving it a spin: the digital magazine Idiomag has an APML application, that would let you focus on the content that you’re most interested in. I’m trying to picture how something like this would play out at a magazine like Pitchforkmedia.com, which publishes thousands of words of content each week, covering a healthy range of musical genres. In a way, that wouldn’t be so different from organizing your RSS feeds to bring you only the blogs and article types that interest you; Idiomag’s trying to do basically the same thing, but from the other direction.

Overall, though, I like this idea. Putting your ten favorite movies and books on Facebook is for chumps. If I could have a copy of my online DNA and meticulously fill it with all of my likes and dislikes, both specific and abstract – likes jittery headache music; hates mid-tempo ballads; likes Bruce Campbell cameos; hates Larry the Cable Guy and all related performers – and then if I could publish the thing on my blog here, and take it to stores and hell, attach it to my writing samples or use it to bolster my pitches – well, hey, that would be pretty cool.

And yes, you skeptics – I think a computer can understand this stuff. I just need a way to teach it.

UPDATE: Here’s another good write-up explaining the in’s and out’s and risks and benefits of APML. As soon as I make some headway in setting up my own file, I’ll post an update.