Pitchformula: music criticism as a creative tool

June 16, 2004 3:51 AM

pitchformula.com This project combines a computer science background and a songwriting hobby with an unhealthy obsession for popular music reviews. In it, I attempt to come up with a new computer-assisted songwriting method which takes music critics' opinions into account. By writing software to statisically analyze the content of several thousand record reviews from the Pitchfork music website (www.pitchforkmedia.com), I generate a set of compositional guidelines based on the musical preferences expressed by the critics. I then use those guidelines to write and record a couple of original songs, discussing in detail the relationships between the songs and the data that I have collected. [via music (for robots)]
posted by soundofsuburbia (18 comments total)

It's every computer scientist's dream: to use their programming prowess to rock out.

Can someone comment on the quality of the songs for those of us at work?
posted by bwerdmuller at 4:07 AM on June 16, 2004

Drum heavy, I thought. Vocals are a little too quiet, in the mixdown.

Otherwise, they're pretty okay. Nothing special. Kinda My Vitriol meets Garbage, imo. Not exactly Radiohead, are they?
posted by armoured-ant at 4:26 AM on June 16, 2004

If this works exactly as it's supposed to, it will produce music that is at best very mediocre.
posted by bingo at 5:07 AM on June 16, 2004

The songs aren't bad, though the percussion is a bit thin.
The vocals remind me of Holcombe Waller, and that's certainly not a bad thing.
posted by grabbingsand at 5:08 AM on June 16, 2004

There are so many cliches to avoid when you're writing about rock, it's nearly impossible not to hit upon a few of them. What makes a rock song work tends to be about texture and contrast, rather melody. And the goal of most indie-ish critics is to expose a band that's barely known, so certain keywords become shorthand for suggesting what a band sounds like, without directly comparing them to well known acts. Over at pitchfork's competition we really strive to avoid the most obvious formulas, though we can get so obscure, I'm sure it turns some folks off. (Disclosure, shameless autopromotion, here)
posted by bendybendy at 5:09 AM on June 16, 2004

The sounds of five years ago... today! Interesting thesis, though.
posted by eddydamascene at 5:20 AM on June 16, 2004

Oh wow! Imagine harnessing all that computer horsepower just to impress a bunch of dicks... kind of like that Final Fantasy movie.
posted by jon_kill at 5:35 AM on June 16, 2004

Now we just need to get Pitchfork reviewers to review these. But, yeah, listenable. It does seem pretty indistinct, but that could just be due to my expecting it to sound so.

Way to miss the point, jon. (I know I know you're just joking etc. but the best satire is based on a true aspect of what it's digging on, not an imagined and unaccurate one.)
posted by kavasa at 6:32 AM on June 16, 2004

The songs are not any worse (or better) than they would have been if they guy had simply looked out the window, strummed his guitar, and kind of made something up in a day-dreamy way. But the idea of actually basing your creative work -- much less your decision of what music to buy -- based on the opinion of music critics is foolhardy to the point of madness. Popular music sucks precisely in proportion to the extent to which it is responsive to or approved by critics. I weep when I think of the catastrophic damage someone like Robert Cristagau has done to popular music over the past 30 years. Even good writers like Lester Bangs are basically full of baloney -- however entertaining in the short run. They bulldoze crap onto the agenda, and it takes years to clear it off, and discover the great, little music that was buried and smothered underneath.
posted by Faze at 6:58 AM on June 16, 2004

Maybe Brent DiCrescenzo will come out of retirement to review these mp3s.
posted by shoepal at 7:04 AM on June 16, 2004

In all honesty, I think Kissing God is pretty good (sounds a bit like a Postal Service demo tape) considering its origins and that rather scares me. It is very much a sonic and lyrical amalgam of many of the records reviewed by pitchfork. Thanks for the link, SoundsofSuburbia. That site is really fascinating and I love that he shares the Perl scripts.
posted by shoepal at 7:19 AM on June 16, 2004

Maybe Brent DiCrescenzo will come out of retirement to review these mp3s.

You could write your own DiCrescenzo review. Just cut and paste 5 or 6 paragraphs from some random livejournal site or maybe some creative writing assignment you did back in college... and presto! If you want, at the very end you can add a line or two about the actual music you're supposed to be reviewing, but it's not totally necessary.
posted by crank at 7:47 AM on June 16, 2004

it's ok, I guess. it's not as good as that new pixies song, Bam Thwok, though. that song is AWESOME!
posted by mcsweetie at 8:00 AM on June 16, 2004

I wonder if this algorithm would work if you plugged in the writings of, say, Richard Meltzer, Nick Tosches, Cameron Crowe and Lester Bangs from the '70s.

(Actually, I'm imagining the result sounding like Patti Smith fronting Blue Cheer. How bad could that be? Well, awful. But how good could it be?)
posted by chicobangs at 8:00 AM on June 16, 2004

The computer part of this seems pretty minor, mostly just analyzing the text of reviews to find words that are scored highly, interpreting what those words might mean and how to produce something that might be described by those words. I mean, he did write all the lyrics, compose all the samples, melodies, notes, play or program all the instruments, etc. Kissing God is not bad, but it ain't great.
posted by RustyBrooks at 8:10 AM on June 16, 2004

my head hurts before listening to the mp3s. i just dont think i can bring myself to download them.
posted by yeahyeahyeahwhoo at 9:38 AM on June 16, 2004

Heh, I came up with the idea of a Pitchfork review generator a while ago. I Googled the phrase and someone else had thought of it too - if I ever make it I'm gonna email him, maybe a collaboration? ;)
posted by abcde at 10:01 AM on June 16, 2004

hey, just saw this article in the nytimes on "biomusicology" :D

[B]iomusicology, is preoccupied with how music affects the brain. What regions of the brain respond to changes in harmony or melody? Is there a single region that makes sense of music? Is there a difference between the way neurons react to frequency differences in speech (intonation) and frequency differences in pitch (melody)? In such research the contingencies of culture and history are often stripped away. The foundations of musical perception are sought, as are the biological laws that make music a human universal...

What sort of picture of musical understanding is taking shape with this renewed interest? Much of the brain research is teasingly inconclusive. Every effort to examine the effects of single musical variables — pitch, meter, harmony — inadvertently shows just how much more music is than the sum of its parts. Despite attempts to identify a particular musical region of the brain, for example, Dr. Tramo has shown that many regions are active when music is heard; even motor areas of the brain can become active though the body might be at rest.

...music has a power unique among forms of human communication: it can teach itself. Gradually over repeated hearings, without the use of a dictionary or any reference to the world outside, music shows how it is to be understood. The listener begins to hear patterns, repeated motifs and changes in meter and realizes that something is happening, that sounds have punctuation, that phrases are being manipulated, transformed and recombined.

Gradually, the listener gains a form of knowledge without ever referring to anything outside the music. Sounds create their own context. They begin to make sense. Similar processes with varying richness and power take place in all forms of music, which is why it is much easier to understand another culture's music than another culture's language.

Nothing else is quite like this self-contained, self-teaching world. Music may be the ultimate self-revealing code; it can be comprehended in a locked room. This is one reason that connections with mathematics are so profound. Though math requires reference to the world, it too proceeds by noting similarities and variations in patterns, in contemplating the structure of abstract systems, in finding the ways its elements are manipulated, connected and transformed.

cheers!
posted by kliuless at 9:54 AM on June 19, 2004