Scholars Scale Up Music Studies

Adam Gerik for The Chronicle

Researchers at the U. of Illinois at Urbana-Champaign and elsewhere are compiling music into a framework that would allow them to study patterns over large collections of data.
May 28, 2010

Finding patterns in music is nothing new. Scholars have studied rhythms, melodies, and harmonies for centuries. Now a project slated for completion in June 2011 wants to achieve something bigger.

An international team of researchers is logging about 350,000 songs—23,000 hours of music—for a new Web-accessible service to give scholars a convenient tool to look into music structures. The researchers hope the data could reveal patterns such as melody extraction and chord description on a wider scope, by looking at classical music of the 19th century, for example, and not just the composer Johannes Brahms.

The Structural Analysis of Large Amounts of Music Information, called the Salami project, is financed in part by a grant administered by the National Endowment for the Humanities. Researchers at several universities are using $1.5-million in grants to create a framework for analyzing musical audio data. They plan to use that framework to process songs acquired from sources such as the Internet Archive and the Database of Recorded American Music.

This isn't the first attempt to build a large-scale database for music, says J. Stephen Downie, an associate professor at the Graduate School of Library and Information Science at the University of Illinois at Urbana-Champaign and a project leader.

But part of what makes the Salami project different, Mr. Downie said, is that it is drawing on expertise from both musicologists and engineers who design databases. That's something previous projects have not done, and it has hurt them, he says.

"The two camps aren't really talking to each other properly," Mr. Downie said. "Engineers come up with excellent programs, but their assumptions about music are naïve. Musicologists tend to be naïve about capabilities of the technologies."

One goal for the service is to encourage increased undergraduate and master's research in music with its convenience and low cost. Eventually the team wants to let music scholars share the algorithms they create by incorporating them into the service.

The project's leaders also want to watch their collection expand beyond its main acquisition now, Western pop music, which is readily available in a digital format.

"A lot of people really build little toys that sort of show the possibility of doing something," says Mr. Downie. "We want to build something utilitarian. We want to move beyond the toy; we want a workhorse."