Wednesday, February 15, 2006

Decoding the Music Genome

In 2000, a group of individuals developed the Music Genome Project, which up to 400 distinct characteristics of a song are captured by trained music analysts. These attributes not only include the musical identity of a song, but many significant qualities that are relevant to understanding the musical preferences of listeners. Librarians call the capturing and assingnment of attributes metadata. The database is built using a methodology that includes the use of a controlled vocabulary, a consistent frame of reference, redundant analysis, and ongoing quality control.

The music genome team has now created an interface called Pandora™ to make the metadatabase available so individuals could use musical 'connective-tissue' to discover new music based on songs or artists they already know. Pandora does not use machine-listening or other forms of automated data extraction.

Enter the name of one of your favorite songs or artists into Pandora and the system scans its database of analyzed music to find songs with interesting musical similarities to your choice. You can then stream the songs over your PC. Licensing restrictions limit the number of songs that one can skip over per hour, so at times you have to listen to songs you do not like. Linkes to Amazon and iTunes provide licensing opportunties.

Users can create streaming "stations" of songs similar to your favorites. Much like with TIVO, you can thumbs-up or thumbs-down a particular song. Stations can also be shared among other Pandora users. Each station gets better the more you tell it what you like, or don't like. Sometimes it doesn't work as one would think. For example, I do not see what common thread Frank Zappa and Billy Joel share.

Hopefully by giving Billy a thumbs down means he will not show up ever again when listening to my Zappa station (he can show up in my 1980's pop music station if he wishes). Sphere: Related Content

No comments: