Music recommendation site listens to what you're listening to

Created by music artist Peter Gabriel, The Filter looks to recommend music, movies, and Web video based on the individual users tastes and mood.

Gabriel and several others have reportedly invested about $5 million USD in The Filter, a service that can apparently recognize about five million songs.

The site describes itself as using a holistic approach to recommendations. At the core is something called Bayesian mathematics, which deals with distribution of probability. In this context, it deals with the relative likelihood that you'll like some particular choice based on the probabilities of other factors that come into play -- whether other people like it, for instance, and whether you like similar things.

Several factors come into play based on what is called an evidence model. The user's purchase and consumption data is combined with his or her browsing habits. Using those as assessment criteria, the site's engine then develops customized recommendations.

"The Filter aims to be the best possible blend of man and machine - a hybrid engine that filters all entertainment content to one's own personal taste", co-founder and chief strategy officer Martin Hopkins said.

Better yet, TheFilter.com also attempts to use your preferences in one type of content, say music, and apply that to its personalized movie recommendations.

Offered by the company is a small plug-in for Mac that works with iTunes, and a plug-in for Windows compatible with iTunes as well as Windows Media Player and Winamp. Once installed, the initial run of the plug-in will do a scan of your music library to see what you own and what you play -- information which is then sent back to the site.

Additionally, during the sign-up process, some initial questions are asked to begin to populate the user's front page, asking for her top three favorite music and movie genres. This is how the process is seeded.

Once that process is complete, the quantity of data provided to The Filter in order for it to make future recommendations is determined by the user. The downloadable plug-in is not necessary -- a user can decide to just use the service's Web site to rate tracks.

Even with the program scanning my library, I still didn't see a change on the front page of Filter. But the company says several matters remain to be worked out, and the site is very much a work in progress.

"We're not going to lie -- the site is public beta -- meaning the site is still in development. We'll still be fixing bugs, making it load faster, improving the design and completing the project for quite some time," the company said in a post on its Web site.