Here’s a neat idea — Echocloud is a service that helps you find new music using a unique approach to building recommendation lists. It works by crawling P2P networks (currently it only supports the Soulseek network, but the developers are planning a Gnutella port), pulling down the lists of files that users are making available for download, and storing those lists in a database. It then builds a correlation model for the database based on the assumption that if two artists appear together in many lists, they must have some commonality that makes an interest in one indicative of an interest in the other. So, if someone searches for one of those artists via Echocloud, they get back a recommendation for the other.

Does it work? Interestingly, the developers say that it works better the more obscure the artist you search on is; this is because obscure artists tend to be in more narrowly focused playlists, while hugely popular artists tend to be found in more generic, “Top 40” style playlists that have less to do with commonality than with ubiquity. To test this, I ran Echocloud searches for two female jazz vocalists, current industry it-girl Norah Jones and the also highly regarded (but less immediately buzzworthy) Diana Krall. The results bore out the developers’ prediction: my Norah Jones search returned mostly links to other high-buzz artists like Coldplay, Madonna, and Avril Lavigne (none of which have much in common with Jones except for high popularity), while my Diana Krall search came back with a much more targeted list of recommendations, including Ella Fitzgerald, Billie Holiday, and Nina Simone. Does this mean their model falls apart for very popular artists? Not necessarily; one could make the argument that, for many fans of Top 40, knowing what everyone else is listening to could be more valuable than exploring more deeply within a genre. For any music fan, though, Echocloud is an interesting idea.