“Herd It” Facebook Game: Aids New “Google for Music” Search Engine

May 17, 2009

(ChattahBox)—A new Facebook music trivia game, named “Herd It,” is not only fun to play, but is also helping a team of electrical engineering doctoral candidates to obtain their PhDs, by feeding the data obtained from the game into algorithms of their new musical search engine the students have dubbed, “Google for Music.”

The experimental music search engine can deliver musical selections to a user searching for a particular genre of music, such as romantic or techno, and the users playing Herd It are helping the students’ search engine to “listen” to and label songs by type, just like a human would.

The Heard It game is entertaining to play regardless of your level of musical knowledge. To play the game, players listen to music clips based on their chosen musical genre and answer questions about the music. The questions range from identifying the genre, instruments used, artist names and providing the correct emotion triggered by the musical selection.

Players score points when their answers match up to other online players. I must in good conscience though, issue a warning to all future players of Herd It: It’s highly addictive!

The team of doctoral candidates from UC San Diego’s Jacobs School of Engineering, include Luke Barrington and Antoni Chan, led by Professor and resident machine learning expert, Gert Lanckriet.

After developing the groundbreaking music search engine, the students realized they needed a more cost effective option to teach the search engine the new vocabulary it needed to successfully “listen” to songs and correctly label them.

The students had been paying; cash strapped UC San Diego undergraduate students to manually label songs to create training algorithms to help their search engine learn about music genres. After running out of cash, the doctoral students came up with the brilliant idea of developing a musical online game, which could help them teach their “Google for Music” search engine.

The data obtained from Herd It will enable the students to train their search engine to find patterns in the songs. The only way for the search engine program to truly learn to correctly label song genres, is to expose it to hundreds of examples of each individual genre, such as romantic, mellow jazz or classical.

Once “Google for Music” is fully trained, songs can be tagged and retrieved based on text-based searches with no human intervention.

Once enough data rolls in from Herd It, the “Google for Music” search engine program can be completed and publicly launched sometime later this year. The data from Herd It will also help student, Luke Barrington obtain his PHD, as part of his dissertation depends on the success of using the data from Herd It to improve the accuracy of the search engine.

So, log on to your Facebook accounts and have some fun Playing Herd It, while increasing your musical knowledge you will also be contributing to the science of artificial intelligence, and helping out a couple of UC students to obtain their PhDs.

The students presented their research at the 2009 International Conference on Acoustics, Speech, and Signal Processing (ICASSP) in Taiwan and it’s available in the ICASSP’s latest journal.

Source


Comments

Got something to say? **Please Note** - Comments may be edited for clarity or obscenity, and all comments are published at the discretion of ChattahBox.com - Comments are the opinions of the individuals leaving them, and not of ChattahBox.com or its partners. - Please do not spam or submit comments that use copyright materials, hearsay or are based on reports where the supposed fact or quote is not a matter of public knowledge are also not permitted.