DATE: SUNDAY, February 8, 1998
SECTION: Sunday Magazine
SOURCE: By John Yemma, Globe Staff
PLAY IT AGAIN, RAM
Christopher Raphael admits that what he is doing is like trying to catch
starlight in a bottle. His quest to teach a computer to perform a musical
accompaniment can be easily mistaken as scientific hubris. But despite his
doctorate in mathematics and his expertise in computer science, Raphael is no
coldhearted numbers cruncher out to back-engineer the most ineffable of the
Muses' gifts. He is a lifelong musician, a concert oboist who has played in
the Santa Cruz Symphony, in California, and is currently a member of the
Pioneer Valley Symphony, in Amherst.
All his life, Raphael has had an intimate connection with music. His
parents, he says, were ``opera fiends.'' He was not a child prodigy, but he
was good enough on the oboe to win a San Francisco Symphony young artist's
competition at age 17.
When he thinks about his favorite piece, Mozart's clarinet quintet,
closes his eyes tight, hearing the sound in his mind, savoring the memory.
Beautiful music, says this 37-year-old professor of mathematics at the
University of Massachusetts in Amherst, ``is a spiritual experience.''
He is not interested in robotizing music, but he is adept enough as a
musician to know what a trap it is to romanticize sound, to believe only in
the beauty of a note or a passage and not the story it is telling. It is
over content, prettiness over intelligence.
And that is why this mathematician who deeply loves music is sitting in
front of a complex string of computer commands on a Sun Microsystems Sparc
Ultra 1 workstation in his office on the 14th floor of a UMass research
``I suspect there will be opposition to this in the music community,''
Raphael says somewhat sheepishly. ``Musicians are very suspicious of
encroachment on their domain. They'll ask, `Why would anyone want to take the
human component out of music?' But I'm trying to make music more accessible
He calls his project ``Music Plus One.'' When he was growing up in
California, Raphael used a popular training record called ``Music Minus
on which, typically, a violin sonata or piano concerto had been recorded
-- once with both the soloist and the accompaniment and once without the
soloist. The idea was that a student would play the missing solo while the
record supplied the accompaniment.
``Music Minus One'' is still around, but, says Raphael, the idea is
fundamentally flawed. The soloist ends up following the accompaniment, which
is, of course, backward. Raphael's aim is to create a computerized
accompaniment that is true, that responds to the human soloist's nuances and
phrasing, that learns from a rehearsal, incorporates the soloist's
interpretations into future renditions -- and through it all remains flexible
to subtle and impromptu changes in a performance. He wants a musician playing
alone to get the benefit of true accompaniment, even if he or she cannot be
part of an ensemble.
You probably think a computer can already do this, given the prolific
of digital sound, the commonplace use of music synthesizers, and the other
wonders we expect from computers on a regular basis. But providing true
musical accompaniment is not an easy task. A computer needs to be operating
real time, listening to a solo, analyzing it, anticipating what will come
next, and creating appropriate music on an electronic keyboard without
missing a beat.
You can hear how well Raphael has mastered this task at his Web site
(http//fafner.math.umass.edu) on Robert Schumann's Pieces in Folk Song Style.
That's Raphael on the oboe, leading the way, the Sun workstation doing a nice
job of staying with him. But not nice enough. The accompanist should not
simply be in sync with the soloist, Raphael says. The accompaniment needs to
make musical sense by itself.
So, for now, he has ``the hood up'' on his program while he tries to
an algorithm that models what a human player would do during a performance.
``There is room for tremendous expression in live performances,'' Raphael
says. ``I'm trying to learn what are the elements that create a feeling of
legato, a feeling of crescendo -- all the musical notions that are not
in physical terms.
``I want to see if a machine can make music that people find
pleasing,'' he says. ``This is part of the broader goal of learning how much
can be accomplished by a machine.''
Mathematicians tend to be drawn to music. In recent years, psychologists
have outlined what they call the ``Mozart effect,'' the salutory connection
between listening to certain types of classical music and performing abstract
and spatial reasoning. Music and math ability, they theorize, reside onthe
same side of the brain.
Though many mathematicians are music fans, few musicians are math fans.
Raphael was not especially interested in math until college, but at
Northwestern University's Conservatory of Music, he found himself drawn to
math courses. Math, oddly, became his salvation, his escape from the tyranny
of musical perfectionism that was beginning to rule his life.
At home, Raphael still labors over the arduous reed-making required of a
true oboist -- the careful selection of the material, and the gouging,
shaping, folding, scraping, and inevitable discarding of most of the reeds
before the right one is found. He practices regularly. ``Never for a
he says, ``have I doubted my deep love of music.''
He loves math, too -- its precision, its rarefied aesthetic. Art needs
science. A good solo needs strong accompaniment.