---

What music will I like?

Issues

The proposed music recommendation service will center its recommendations around each listener, with the understanding that each listener interprets and understands music differently. The service will understand that no one listener has inherently better taste than another; however, a listener's preferences might correlate strongly to those of another user, and therefore his recommendations are likely to prove more valuable. The service will aim to provide these recommendations in a simple and accessible fashion and provide basic insight to the casual music fan, yet also generate statistically accurate recommendations in a technical and precise manner to avid music listeners.

Methods

Because of the central nature of listeners in the music listening experience, the key to the proposed service will be finding reviewers (professional, blog, or peer) with similar preferences, and therefore similar understanding. The system will then deriver recommendations from a weighting system that gives significantly greater weight to similar listeners and significantly devalues dissimilar users. When first joining the service, the user will be assumed to have a totally average taste in music. As the service collects various inputs (ratings, tags, and playcounts) from the user regarding his/her listening preferences, the service will scale its "average listener assumption". As the service collects more and more information, the amount by which the service will scale recommendations, and thereby the confidence the system will have in its recommendations, will quickly increase. Unpersonalized similar music networks will also be generated to help listeners quickly understand about the music to which they are listening. Although man-powered audio analysis is inefficient and expensive, research will be conducted into computerized audio analysis techniques and as these techniques are proven to be effective, they will be incorporated and appropriately weighted into the personalized music recommendation engine.