February 13, 2012 at 1:36 pm

Music Hack Automatically Predicted Grammy Wins Almost as Well as Billboard.com

paul vs billboardOf all the hacks cobbled together at Music Hack Day San Francisco this past weekend, none were more timely than Paul Lamere’s, which attempted to predict Grammy Awards winners in 13 categories automatically, by analyzing what people were saying about the artists in question.

Just hours before last night’s ceremony, Lamere revealed that his app — created in 24 hours — predicted the winners nearly as well as Billboard.com did [updated: Billboard points out to Evolver.fm that the article in question was based on the informed opinions of Billboard.com's tastemakers and did not include Billboard sales data].

As well as being an active member of the music hacker community, Lamere heads up the developer platform for The Echo Nest, a central organizer of Music Hack Days (as well as publisher of Evolver.fm, an independent publication). His hack seeks to prove that computer algorithms can gauge the web buzz surrounding artists to predict things like the outcome of the 2012 Grammy Awards.

Billboard has been a source for music data since the launch of its first “hit parade” chart in 1936, and forges its yearly predictions on a detailed understanding of digital and physical sales data, as well as news and the  industry’s general impression of the artists [Update: Again, Billboard's article was not meant to reflect that data, but rather the Billboard.com staff's expert opinion].

The Paul vs. Billboard hack takes a more automated approach, leveraging The Echo Nest’s contextual data to predict the winners based on the level of internet buzz from blogs and other publications in the days leading up to the awards. Lamere’s machine-based method of prediction would prove its effectiveness later that night.

With a final score of 6-7, Paul vs. Billboard barely trailed Billboard in accuracy. The app made better predictions for Song of the Year, Best New Artist, and Best Pop Solo while tying Billboard in three categories (both predicted Adele’s 21 for Album and Record of the Year, and Bon Iver [a.k.a. Bonnie Bear] for Best Alternative Album).

Lamere’s updated post compares his and Billboard’s predictions to the final outcomes. While he concedes that man barely bested machine this time around, Paul vs. Billboard offers an encouraging look at how machines can learn to understand humans — and the music they will likely reward.