Saturday, January 5, 2008

In the mood.

We build playlists because we tend to want to listen to music to suit a particular mood. That's why we all have at least one CD (or tape, if you're my vintage) in a drawer somewhere with a hand-scrawled label reading "80's Freak-Out" with someone's carefully chosen collection of songs that just scream, well, "80's Freak-Out".

Of course, in this age of digital music it's easier than ever to create playlists to suit our whims. No more cueing up LP's and deftly pressing Record and Play. With just a bit of drag-and-drop anyone can cull together a bunch of similar songs to provide just right ambience for the occasion.

Unfortunately, just because the mechanics of building playlists has gotten easier doesn't mean that picking the music for those playlists has progressed beyond the earliest days of time-shifted entertainment media. I can just imagine Thomas Edison's internal dialog: "Okay, everyone has aperitifs. Now for some light party music on the phonograph, but which cylinder? 'Hunting Horn and Gun-blast'? No, no, too edgy. Perhaps 'Barking Dog' or 'Train Whistle'. Think, Tom, think…"

True, we don't have to sort through stacks of LPs or piles of cassettes to find the song we're looking for. But now we've got hundreds or thousands of hours of digital media all neatly tucked away on our big, cheap hard-drives, with the perfect next song for our "Depression Session" playlist no more obvious than in the old days. Sure, some lucky few possess the razor-sharp memory and encyclopedic knowledge of recorded music that allows them to effortlessly pluck tracks out of thin air for any mood. Unfortunately, most of us aren't that person.

That's why since nearly the beginning there have been DJ's and program directors and the like who actually get paid to do that sort of thing full-time, ostensibly because they *are* that person. But the explosion of accessibility to new, interesting and obscure music thanks to the Internet has finally confirmed something we've suspected all along. Even the best DJ's and professional music geeks are only passable at picking what we want to listen to. At it's best it might be an unformatted college radio station that's introduced you to some really cool, undiscovered stuff over the years, and at it's worst it's the relentless, every-hour-on-the-hour, hyper-target-market-formatted, heavy-rotation, 30-seconds-of-ads-every-eight-minutes, I-can't-believe-they're-playing-this-again, top-40 nightmare. In the end, it comes down to the basic fact that our tastes and preferences are complex and ever-changing. No form of old-fashioned play-listing is really going to deliver a peak listening experience with any sort of consistency.

The promise of digital media has always been wrapped up in the ability to tie meaningful meta-data to that media and cross-correlate it in new and novel ways to (hopefully) bring us closer to the media and provide us with a richer, more enlightened listening experience. Unfortunately, for a number of reasons, that dream has usually come up woefully short in practice.

One problem is the quality of the meta-data. Ever since the earliest days of CDDB we've struggled with sloppy, incomplete, questionable, or just plain incorrect meta-data supplied by some anonymous yahoo on the 'net with too much spare time. Another problem is the structure of the meta-data itself. Sure it's nice that sometimes the "Year" field is populated (and maybe even correct) or that someone thoughtfully applied what they thought was a meaningful Genre to a track. But just how useful is that information? Maybe you could build an "Uber-80's" playlist based on the Year field, but would it really capture the essence of the 80's in a couple hours worth of music, or would it be 287 songs with no more in common than the date they were released? How do you ensure that your "Uber-80's" playlist includes the essential "The Chauffer" by Duran Duran and definitely does NOT contain Billy Ocean's "Caribbean Queen" (or the other way around)? With a little work you might be able to ensure everything works out by rating all the songs in your library (I have a suspicion, but no evidence, that song ratings are the most common way that people filter their music). But ratings alone don't really solve the problem because "Caribbean Queen" might be a "2" for your normal listening tastes, but an absolute "5" for your "80's Songs You Will Never Admit to Loving" playlist. After all, it's gotta be at least a "3" on some dimension, or why even bother having it in your library?

There is today a preponderance of niche internet radio stations and new, fancy services like Pandora that try to guess what music you might like. But none of this contributes any value to the fundamental pursuit of spontaneously selecting "mood-appropriate" listening material from your existing music library.

So where does all this leave us when the new squeeze is coming over to our pad for the first time, or we're hosting a little cocktail thing in the private room of the hip bar down the street, sans DJ? Right back where we started, unfortunately.

The solution, I think, is the simple, powerful concept of tagging. Arbitrary, contextually meaningful chunks of descriptive text that you, the listener, have chosen to associate with a particular track. The concept of tagging has become commonplace in the age of Web 2.0, with just about any on-line activity being augmented with some sort of "tag -this" functionality. But for the gigs of music sitting right there on your hard disk? I searched and couldn't find anything. So I did the only thing I could think of, I wrote my own little tool, called Tweet, that does exactly this. It's built on top of Apple's iTunes, my music player of choice. If you're an iTunes user and you want to be able to easily tag your tracks (and aren't afraid of beta software) then you might want to check it out. It's free. And by all means, if you do, please let me know what you think.

Add to del.icio.usDiggIt!RedditStumble ThisAdd to Google BookmarksAdd to Yahoo MyWebAdd to Technorati FavesSlashdot it