Collaborative filtering for music picks up pace

By

I believe that “collaborative filtering” is at the heart of how the networks are coming to life. The basic concept is that we must collaborate to filter the massive information overload we face. We can do this simply by taking the recommendations of friends we know and trust. However now software can amalgamate the views and perspectives of thousands or millions of people to direct you to the information or entertainment that is uniquely relevant to you. One of the many domains in which collaborative filtering can be applied is music, by recommending or playing music based on your preferences.

Last.FM, which I last wrote about in this blog in 2003, has developed a lot further, and seems to have gained critical mass. Last.FM’s underlying recommendation engine, Autoscrobbler, is now also available separately, helping to fund Last.FM’s free, no-advertising offering. Last.FM creates personalized radio stations based on people’s preferences, and makes recommendations based on what other people with similar tastes like. As such, it is intrinsically based on social networks, and also provides group functionality so people with similar taste can interact. A newer service, Pandora, works quite differently. It uses sophisticated algorithms to analyze music based on its rhythmic, melodic, and harmonic structure, instrumentation, production qualities, lyrics and so on. You seed the service by nominating a song or an artist, and it creates an entire radio station from that. You can also provide feedback on your preferences, as in Last.FM. The service is available for free with advertising, or there is an advertising-free subscription service. Pandora raised $12 million last November, bringing its total raised to over $22 million, showing the faith investors have in the value of these models. However one venture capitalist prefers Last.FM to Pandora in an interesting comparison of the two services. Check out these music services and support them! The better they get, the more we can discover and listen to the music we love the best from the many millions of songs produced every year.