Comments on the Music Genome Project
For those who don’t know, the Music Genome Project is the database behind Pandora‘s streaming service. Unlike many services that follow the recommendation format of “you like song ‘a’, and Bob likes song ‘a’ and song ‘b’, so you are likely to enjoy song ‘b’ as well”, Pandora’s algorithm is “you like song ‘a’, and song ‘b’ sounds like song ‘a’, so you will probably like song ‘b’ as well”. This method, at least to me, seems to be much more accurate.
Recently, while searching for API documentation for the Music Genome Project, I have come to the conclusion that Pandora does not have a public API for their service. If I am incorrect in this and just cannot figure out how to use the Internet, please let me know. If, however, I am correct in my findings, I would strongly encourage Pandora to release their API to the public. I can think of many uses for this API outside of making a competing music player. Two projects that I would love to create – if I could find the API documentation – are listed below.
The first project that I would like to create is a mash-up between the Music Genome Project and Amazon. Lets say you have a friend who’s birthday is coming up soon, and they are a huge Beatles fan. You don’t want to risk buying them a CD that they already have, but what can you do? Easy, enter a band into the mash-up application, and get back a list of albums on Amazon that would likely appeal to your friend.
The second project that I am considering is a service where people can create an account and list all of the groups and songs that they like. The site would then find matches to music that is similar to the user’s favorites and provide a list of fellow users who enjoy these similar artists and tracks. Users could then ask people in the provided list questions about the band (i.e. favorite song or album from the group). Basically, this would be a music-oriented social networking site geared towards a community of users interested in giving and receiving personalized information about artists instead of reading a review intended for a broad audience.
However, without a public API, both of these projects will just have to wait. Alternatively, there is a group working on an open version of the Music Genome Project. A fully functional release of this service seems to be pretty far off though.