Bought the latest Coldplay album and wonder what else you might like? Four German companies are competing to offer the best online music-recommendation software. In the mix: the inventor of the MP3.
You like Marilyn Manson? Maybe you'll like Cher as well
Some time in the near future, Karlheinz Brandenburg would like to have a conversation with his stereo.
He'd like to tell it to play the kind of music he likes to typically hear at that time of day and what music direction it could consider.
"That’s a vision I’ve had for four or five years," he said.
For the inventor of MP3 technology (photo), the preferences would lean towards Bach and Beethoven.
Karlheinz Brandenburg, one of the inventors of the MP3
But new software Brandenburg and his colleague Markus Cremer have developed at Germany’s renowned semi-public think tank, the Fraunhofer Institute, will enable his stereo to guide him to a 19th century string quartet he’s never heard of -- or maybe even some Björk.
In the cut-throat world of music downloading sites, the survivors will be those that offer the best service, said Brandenburg. One option aimed at adding value to a music download site: music-recommendation systems that link users to songs that fit their taste, but that they might otherwise never have come across.
Like Marilyn Manson? You'll love Cher!
"The task of a piece of software to help you on one hand to make it easier to find things, on the other hand to find the unexpected," said Brandenburg in an interview with DW-WORLD. "To find things you like even if they weren’t in the same catalogue or next to each other."
At present, four German companies are competing for the chance to help consumers match music to their tastes. Brandenburg and Cremer think the answer lies in computer software that analyses the music itself. Their program, called Soundslike, measures song ingredients such as rhythm, melody and instrumentalization, then recommends songs that have similar patterns.
That leads to tips like Billy Idol’s “Cradle of Love” for users buying the Rolling Stones’ “Sympathy for the Devil.” But it can also lead to a recommendation of Cher’s (left) “Alive Again” to buyers of Marilyn Manson’s version of “Tainted Love.”
“I wouldn’t rule out that there are some kinks to work out,” said Brandenburg, chuckling.
Brandenburg knows all about working out the kinks. After all, years passed between the first time Brandenburg compressed a digital music file of Susanne Vega’s “Uncle Tom’s Diner,” in the mid-1980s, to the point in the mid-1990s when his MP3 technology began to take off. He thinks Soundslike technology could be ready for a company to implement in a few months.
The human touch
But his competitors say the program has fundamental flaws. It's not the music that should be measured, but rather the consumer's taste.
“It’s not possible for a machine to differentiate between a piece of grunge music or heavy metal from the 1980s, because it looks very similar acoustically,” Daniel Müllensiefen, project manager at Musicline.de told DW-WORLD.
His system, Soundprofiler, features a database compiled through a mix of human suggestions and software. After rating a list of 40 songs, ten at a time, users are given a final list that is supposed to match their tastes perfectly.
Beginning at the Popkomm music trade fair in Berlin on Sept. 29, Internet users worldwide will be allowed to test the two programs. They can then decide which of them -- or one of two others designed by Hamburg-based HIFIND and software developer DDD Systems -- works the best.
The winner gets an exclusive partnership with Phonoline, which provides the software to Germany’s music download portals. Along with the Internet voting, a jury of music and technology journalists will help decide who gets the contract.
iTunes cool to the idea
But whether big names like Apple or Microsoft will bite is still unclear. Alex Luke, director of music programming and label relations at iTunes, said that Apple continues to recommend songs based on what buyers have bought and what they suggest.
“How users buy and build their collections is valuable, but it's only one of the ways we recommend music on iTunes,” he wrote in an e-mail.
But buyer statistics and suggestions don’t go far enough, argues Brandenburg.
"There you have the problem that ... you always get to the same mainstream music," he said. "We clearly didn’t want that to happen. You need to have something unexpected in there."