samedi, juin 24, 2006

Music for Everyone

Music search sites that learn your taste

By Adam Pasick Fri Jun 23, 4:51 PM ET

LONDON (Reuters) - The boom in digital music has left a lot of iPods to fill and made millions of songs available online. But how to choose which ones? Music recommendation services are using tech wizardry to solve the problem.

It used to be so simple: Music fans discovered new songs by talking to friends or listening to the radio, then paid a visit to the local record store. But now, with online music stores like iTunes and Napster offering millions of often obscure songs, users are searching for a better way.

Two of the most popular services, Pandora and Last.fm, take radically different approaches.

Pandora ( http://www.pandora.com ) sets up a personalized online radio station based on a few favorite artists or songs, then adds new songs basing the selections on attributes of the music you've chosen.

For example, I set up a Pandora station of soul music ( http://tinyurl.com/ltu7a ) with artists like Aretha Franklin and Donny Hathaway, which I then refined by giving a thumbs-up or thumbs-down to various songs while listening on my computer.

How are the songs selected? According to Pandora, I like songs that "feature mixed acoustic and electric instrumentation, electric pianos, and subtle use of strings," among other things.

The idea that my musical taste can be so easily pigeonholed is slightly distressing, but the results are uncannily accurate.

THE WISDOM OF CROWDS


If Pandora recommends songs based on their inherent qualities, Last.fm ( http://www.last.fm ) takes a very different approach by relying solely on the power of social networks: If you and a lot of other music fans like one song, it's likely that you also have other favorite songs in common.

The concept is known as collaborative filtering, and it often shows up on Web sites like Amazon that offer recommendations stating that "customers who liked X also liked Y."

Last.fm works through two separate pieces of software: One that monitors the music you listen to on software like Apple's iTunes and another that streams a personalized radio station to your computer.

The Last.fm Web site also allows you to search for a given artist and find similar music, as well as listen to 30-second samples of most songs.

The statistics on my listening choices were surprising -- Do I really listen to the Beastie Boys that much? -- but I liked most of songs the service delivered.
Last.fm also has grown into a large online community of music lovers, based on shared musical taste.

TO EACH THEIR OWN


Last.fm and Pandora are far from the only services vying to separate the musical wheat from the chaff. Other sites include Live365 ( http://www.live365.com ), which offers some 7,500 user-generated radio stations, and MusicStrands ( http://www.musicstrands.com/ ), a comprehensive music recommendation and community Web site.

Live Plasma ( http://www.liveplasma.com ) is a site that draws striking graphical maps that show the overlapping relationships between artists. Aretha Franklin, for example, is orbited by Teddy Pendergrass, Chaka Khan, Marvin Gaye and Barry White.

Despite the use of technology, musical taste remains an idiosyncratic matter for most people, so it pays to try multiple services -- and to remember tried-and-true real-world methods.

Mitch McAlister, a 30-year-old American living in London, has used Last.fm and Pandora with varying results.

"Last.fm is somewhat interesting to me, but I think Pandora kinda sucks all around -- for some reason they keep recommending Coldplay for all my music choices," he said. "Last.fm has been measuring my iTunes plays for a while now. I think there might be too many options and too many different applications, but I don't get Coldplay over and over again."

In the end, McConnell said that neither service exposed him to new music he truly liked. For that, he relies on methods that involve little to no technology at all.

"I am a cynic when it comes to music. I go to shows and talk to people," he said.