Re: robot com-posers

Subject: Re: robot com-posers
From: Eliot Handelman (
Date: Thu Feb 17 2005 - 19:10:48 EST wrote:

>Sure, I know you're already working on it :) Good plan making people think it
>was a useless idea...
 We've already seen a few posts inspired by your suggestion. In fact,
very many people,
especially those not technically knowledgeable about music, think that
all you have to do is code
up the rules of music and there you go. The trouble is, we may sort of
hear those ruoles in
action, but we don't know how to extrapolate them and generalize the
circumstances in which
they can be expressive. You can't just borrow algorithms and get at
some sort of stylistic
grammar even of smple miusic. Even if you had one, are you sure -- to
borrow Miriam's
expression -- that's what makes the music tick?

The less you know about music the easier this seems. Why we've all
heard of the fact that
a sonata has a coda, for instance. It must be easy!

Surely theorist's have mapped it all outy by now, said someone to me
recently. It's noit the
case though.

There was a paper out a few years ago in which the author had had a
thunderbolt iuntuition:
that you can use regexp matching on midi data. Of course, he'd never
really thought much about music
and all he could do was find literal matches, without being able to
figure out why we're able to detect
variants of tunes and such. The thunderbolt wilted into a flaccid old
shaft of lettuce.

More papers were written about "string matching with gaps." The idea
here is, if some notes are the same,
then maybe we found a variation? He didn't say what kind of variation.

Many people are currently working on such problems, amazingly without
the least theortical insight into what i
it is they may be looking for, and of course they're finding nothing.

Neverthless, this is where the bulk of research is, justifying your
paper's opening statement:

"Much interest has currently been expressed in ...."

Why DARPA may even fund you.


>ps: the main point was to focus on 1 type of music rather than "MUSIC". I'm
>pretty sure that would simplify things for you. And the database of music...
>you gotta have that too... anyway don't listen to me I'm just a student...

You're right, though -- you do need a database, and you should
concentrate on something fairly specific.
There's an excellent collection of german folksongs available that I've
worked with
for a few years. I also spent a lot of time on Chopin. I've been
thinking about moving to raga for a while. I find that there are very
things that happen in all music -- not in the sense of note or rhythm
pattern, but in the sense of parallels, the ideas of rhyming,
on parallelized motion and opposition, and rhythmic energy, and surely
of renewal.

-- eliot


This archive was generated by hypermail 2b27 : Sat Dec 22 2007 - 01:46:06 EST