Lehigh Week, 18 October 2000

Pianist and engineering professor, she marries music and computers

Endless hours of scales, arpeggios and technical studies can make a pianist wonder:
      When does the jumble of notes become music? When does preparation yield to creativity? When, possibly, does talent reveal a glimpse of the sublime?
      Elaine Chew goes a step further.
      As a concert pianist, she strives to paint a canvas, to tell a story, to convey emotion, passion, history. Making music happens so quickly, she says, "that it almost becomes an unthinking action."
      As a practitioner of artificial intelligence, Chew creates mathematical models to teach computers the most rudimentary musical concepts. "What is a key, or stable pitch?" she asks her models. "How do you know when a key has modulated? How do you know when chords are shifting?
      "A mathematical model imitates what the human mind is able to do," says Chew. "But it does so poorly. It is sort of an impoverished mind."
      Even a musically untrained human ear can detect a change from highly structured Baroque or Classical-era music to the less-restricted Romantic style, Chew says.
      "But artificial intelligence can't. What is different about the mind? Why can it learn and adjust?
      Chew, a visiting assistant professor of industrial and manufacturing systems engineering, will give a lecture-recital titled "Impressions" at 4 p.m. on Wednesday, Oct. 25, in the Zoellner Arts Center. She will perform Ravel's "aspard de la Nuit" and introduce local audiences to 20th century Chinese composers Jiang Weny and Wang Lisan (http://web.mit.edu/eniale/ChineseMusic/concert). Chew grew up in Singapore loving both math -- her parents are math teachers -- and piano -- she started lessons at age 6 and made a name for herself as a soloist and accompanist.
      When she was 19, Chew enrolled at Stanford, where, in three years, she completed a bachelor's degree in arts (music) and science (mathematical and computational sciences). Then she went to M.I.T., earned an M.S. and a Ph.D. in operations research, wrote a thesis on "A mathematical model for tonality" and found time to spend a summer in China searching for piano and chamber music written by 20th-century Chinese composers. She returned with several boxes.
      Using themes from Book One of Bach's Well-Tempered Klavier, Chew created a mathematical model that could detect the key of a melody faster than could other models.
      "Music is so complex that it's very difficult to model," says Chew, "but in making the effort, we can learn more about how to model other different things."
      "Music cognition is pattern recognition at multiple levels. I try to connect spatial construction with a part of music, to relate spatial representations with tonality in Western classical music."
      "The computational aspects of computer science and artificial intelligence are very interesting, but even more interesting is the human aspect. What do human beings do? How do I know what to stress, how to interpret, where to put an inflection? How does the brain sort out information and make something sensible?
      "It's like the brain sorting out information in any field -- it's akin to how do we make sense of the world we live in."

- Kurt Pfitzer
Photo by John Kish IV