Re: SARA

>From: Sarn Richard Ursell <thcdelta@pop.ihug.co.nz>
>
>-This is not not the actual
>article I was referring to, the one I DID refer to contained a MUCH more
>detailed description of SARA, (it was int New Scientist I believe)....
>I have heard algrotihmic music (genetic, stochastic, random, Brownian,
>Mandelbrot set generated), however all of these sound good, and interesting
>allright-but kind of phony.

SARA and the ideas behind it are described in Cope's Book _Experiments in
Musical Intelligence_ (Madison, Wis.: A-R Editions, 1996). It comes with a
CD-ROM with SARA and a lot of example databases, though you need MCL Lisp
to run it.

The very brief idea behind SARA, as Cope explains it in the book, is a kind
of super-generalization of the 18th century musical dice game, where one
could assemble realistic-sounding piano pieces by choosing from a list of
possible first measures, second measures, and so on. So all the possible
first measures would be in the tonic, the fourths a half-cadence, the
eighths a full cadence, and so on.

SARA takes the same idea that some musical fragments can be analyzed as
"fitting" hierarchically with others, based on things like melodic
direction, the fragment's "significance" as a distinctive marker of the
style, and a kind of Schenkerian reduction of the fragment's place in an
overall phrase and form. (Naturally this process works best with music that
can be successfully analyzed in this way.)

SARA is not so simple that you just pour in a lot of pieces and it starts
spewing out variations on them. You have to select the pieces very
carefully, break it down into fragments (choosing the right length for
fragments is key), do all the analysis, and carefully choose from the
output.

It is a very subjective process, which is something most press accounts
have glossed over if they mentioned it at all. People are most astonished
that a computer can "compose" music in the style of Mozart (or whomever).
Personally, I think that the computer is mainly rearranging (very
intelligently, of course) notes that Mozart wrote based on David Cope's
analysis of Mozart's style. Even then, Cope throws away many pages of
output for every one he retains.

Could SARA work for non-12TET pieces? Well, it works pretty well for Bach
and Mozart. It does treat all enharmonics as equivalent, though, if I
remember right. If the tuning would work in 12 or fewer pitches per octave,
I don't see why it wouldn't work, but SARA doesn't really know anything
about the tuning. It deals with note numbers and rhythms, not what
harmonies sound like.

To be most useful, SARA would normally have a database of fragments from a
large number of closely related pieces, transposed to the same tonic (if
tonal) and in the same meter. This is not to say that interesting results
could not be found by deliberately creating "hybrids" by mixing composers
or styles. Cope has done a little of that, mixing Stravinsky and Mahler for
example.

If you had a large database compiled of a composer's 10TET pieces, I
imagine it would work fine. Any logic to the 10TET harmonies would probably
be retained and reproduced in the same way that it recreates the logic
behind Bach chorales. I don't know of any composer who has written a lot of
10TET pieces, though (much less in the same meter and general style).

Best of luck.

Bill

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^ Bill Alves email: alves@hmc.edu ^
^ Harvey Mudd College URL: http://www2.hmc.edu/~alves/ ^
^ 301 E. Twelfth St. (909)607-4170 (office) ^
^ Claremont CA 91711 USA (909)607-7600 (fax) ^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^