After catching up on your roommate’s life through his or her wall posts, Facebook stalking the girl who just loaded scandalous new party pictures and updating your status for the third time in less than 24 hours, you can now add an application to keep yourself amused during lectures ‘mdash; and be a part of a new developing music analysis technology created by UCSD’s engineering department.
Herd It, an interactive game that allows players to rate a song’s danceability, intensity and likeability; awards for common answers and the ability to identify a song’s title and artist.
‘It’s called ‘Herd It’ because we’re all listening to music but we’re also part of this big herd,’ said Luke Barrington, a Ph.D. student in electrical engineering. ‘We’re listening to these same songs.’
Barrington has been developing the application’s technology with Professor Gert Lanckriet of the electrical engineering department.
User answers form a data set that will grade, for example, the ‘pleasantness’ of a Beach Boys song, and will be used to program new technology that will be able to define pleasantness on its own by similar wave patterns between songs.
After the program is trained to identify musical qualities, the data set ‘mdash; made up of thousands of songs ‘mdash; will be navigable with a search engine that responds to real and descriptive language.
‘If I’m looking for scary Halloween music or if I want to listen to some aggressive rap tonight, I can type that in and find it,’ Lanckriet said.
Barrington and Lanckriet hope their search engine will work faster and more efficiently than similar systems offered by Pandora or iTunes Genius, which require the user to type in a song title or artist.
In order for the program to recognize which songs are ‘upbeat,’ ‘romantic’ or good for ‘doing chores around the house,’ it has to be supplied with example wave patterns. Lanckriet and Barrington built this database by paying UCSD undergraduate students to describe 500 songs with adjectives, the way Pandora categorizes its music.
So far, the program contains a limited amount of musical knowledge, and is operating on a beta version of the search engine. However, the pair’s final goal is to retrain the computer with the data collected from Facebook’s Herd It to categorize an unlimited amount of music.
‘The whole motivation for this project has been the fact that there are all these independent musicians out there,’ Lanckriet said. ‘It’s challenging for these independent artists to get their music to access a lot of people, and since myself and my students are also musicians, we felt that pain.’
The search engine achieves this by providing music without artists or song names, which are necessary to create a Pandora station. I
t could also become a platform for musicians to broadcast their music to the industry, including recording label producers, advertising agencies and movie producers.
‘For example, I’m making a BMW ad and I’m looking for some young, sporty, aggressive driving music and I don’t want to pay $20,000 or $200,000 to have some well-known band give us a song,’ Lanckriet said. ‘What about the guy that has a synthesizer sitting in his attic and actually makes some really cool music and is willing to sell it for a couple hundred dollars?’
The goal of the musical-analysis program is to connect those looking for music with those who have it, he said. But an exact business plan for turning it into a commodity is still unclear. At this time, the engineering department is merely exploring a technology based on academic research and rooted in common interest.
Lanckriet and Barrington said they hope to launch the technology to the public this summer, but added that their invention will not achieve perfection.
‘Our experience so far is that there is no substitute for a trained human ear,’ said Tim Westergren, the founder of Pandora Radio, to VoiceofSanDiego.org.
The successful virtual categorization of music wave patterns, according to Barrington, also suggests that future programs could create music with these descriptions woven into their database already, regardless of whether the song is actually on key or subjectively ‘good.’
Though Lanckriet and Barrington’s invention would potentially give a stage to undiscovered talent, it could also expose songs under false pretenses.
‘It’s a good opportunity for independent artists, but at the same time, it’s flooding the market,’ said Kari Francis, a sophomore majoring in music at UCSD. ‘It may be difficult to differentiate between good talent and people who have access to the technology. There are so many people who have GarageBand and all these programs now that make producing music really easy. People are able to use the computer as a tool to manipulate their talents and skills in such a way that they may not actually be as talented or as skilled as what they’re producing.’
Readers can contact Ashley Lee at [email protected].