Hackathon: Can One Million Fan Interviews Recommend Music?

Guest post by Eliot Van Buskirk of Evolver.fm.

There are many ways to skin the music recommendation cat, so to speak. Record label EMI and Data Science London hope they have stumbled on to a new one: data sliced -and diced from interviews with one million music fans.

The team behind the EMI Million Interview Dataset will release the self-reported musical taste of over one million music fans worldwide this weekend, so that developers and hackers at this weekend’s Music Data Science Hackathon can try to build musical recommendation apps based on what those music fans said about the music they like.

In addition to EMI’s interviews, conducted by Lightspeed Research, the hackathon will incorporate technology from EMC, “a world leader in data science and big data solutions” and Kaggle’s “collaborative, real-time, online platform for predictive modeling competitions.” Adatis, a UK-based consultancy, will offer a £500 “data visualization prize” at the event, which includes £6,000-worth of other prizes. If you’re looking for the dataset you won’t find it today, but it should be available here starting this weekend.

The big test: Can these interviews enable a machine to recommend a new song to a listener based on what they said in an interview — demographics, word associations, and their similarity to past interviewees? We should know the answer — or rather, the initial answer — after the 24-hour Music Data Science Hackathon (London, July 21-22).

“With the EMI Million Interview Dataset we hope to bring more new ways of thinking into our industry that will deliver enormous benefits to artists and their fans,” said EMI senior vice president of insight David Boyle.

By mining these million-plus interviews, EMI and the other sponsors of the event hope to produce a number of creations that can predict whether a given person will like a new song they have not heard before.

Plenty of other folks are working on these issues, of course. Being a record label, EMI hopes to use the ability to predict musical taste in order to “deliver enormous benefits to artists and their fans.”

Photo courtesy of Flickr/Laurence Barnes