Sunday, October 22, 2006

The songs that make the whole world sing

I just read this article in The New Yorker about how neural networks are being user to analyze music and movies to predict whether or not they are commercially viable. Most of the article is about analyzing movie scripts to predict box office numbers but there's a bit at the beginning about a company that uses mathematical equations to analyze the hit potential of a song. Since music really can be broke down to waves which can be broken down to math (the company makes no effort to evaluate lyrics) this seems more doable than analyzing movies. And, to a former mathlete, this is pretty intriguing stuff.

The article barely touches on the potential controversy of trying to pin down art with science but I can't imagine that many people wouldn't find the idea that a "good" song could be constructed in such a callused manner fairly upsetting. Music is supposed to be about self expression and it may seem that the soul of a song is somehow lost if someone tweaks with the bass and the rhythm in an effort to hit the popularity sweet spot. My initial reaction to the idea of constructed music was to distance myself, thinking, "sure, they can do that with music for the masses but not to me! I have taste! I don't listen to the slop that's popular!" I even felt that if I did like one of these songs I was somehow being made into a fool.

It's fairly accepted that the human idea of what is physically attractive in a sexual partner is heavily influenced by our biology. We all subconsciously look for partners with good genes who will help us to successfully carry on our own genetic lines. We look for a partner with symmetrical features and no clear signs of disability. This does not diminish the idea of love. Knowing that some of my attraction to another person is animalistic does not make my feelings less real. It seems that the same logic would indicate that we find certain sounds pleasing because our ears have been programmed to seek them out. If my ears are waiting to hear a certain collection of notes I don't think it matters if that collection is created purposefully or by accident.

The more I think about this the more I really wish I had access to this software so I could run my cd collection through it. The variables that make up a song (ratio of high to low notes, pitch, etc) all boil down to a point on a graph, when you overlay the graphs for other well known songs you see that popular songs fall into one of a few "hit clusters." I'm dying to know what hit cluster my personal soundtrack falls into. Think of how a personal music analysis could revolutionize online dating for geeks like me? If they could create book, TV and video game versions and superimpose those graphs I would never need to read another long winded autobiography (more importantly I'd never have to write one). The only things techno hipsters care about anymore (besides uhhh symmetry) is pop culture, if we could create visually pleasing easy to read pop culture calling cards we could all stop wading through the sludge on myspace. That would be one the greatest scientific accomplishments ever.


alia said...


i heard something (WNYC, of course...) where they had a computer analyze this one particular baroque composer. he wasn't particularly famous (um, to non-classical-music experts, anyway)... kind of a systematic composer, so it was an easier fit. anyway, what they found was that yes, the computer followed the musical rules expertly, and the music was "good".

the interesting part was the lyrics: in the original score, the music is sad because the lyrics are sad. and when the words said something about "soaring" the music notes went up.

but since the software had no idea about the meaning of the words, it was just matching them up with notes, it had this really playful, fun melody to heartbreaking words.

which on the one hand was odd, but on the other-- there are some famous pop/indie songwriters who have done that... but then i think "well, it's art because they *meant* it..."

but then i just get confused...



Ronan Jimson said...
This comment has been removed by a blog administrator.