Briefly Noted

Music, it has often been suggested, is a form of language. If this is true, perhaps some of the techniques of computational linguistics might be profitably applied to musical issues. In Computers and Musical Style, composer David Cope describes a new way of analyzing musical style, by using computer pattern matching to identify characteristic motives, or short sequences of musical intervals. Cope has developed an automatic composer, using an ATN music grammar to recombine these motives, producing convincing music in the style of composers ranging from Mozart to Gershwin. Cope's primary motivation in this work was a practical one: suffering from "composer's block," he felt the need for a "composer's partner." To that end, Cope has developed a series of computer programs capable of composing music in the style of a particular composer. Drawing on the frequently observed parallel between linguistic phrase structure rules and the structure underlying musical phrases, Cope defines an ATN grammar as the basis of his automatic composer. Using work by music theorists such as Schenker, Cope defines five musical categories: statement, preparation, extension, antecedent, and consequent. The grammar governs the ways in which these five basic categories can be combined to form phrases. Cope creates a lexicon by extracting characteristic patterns from existing works of a given composer, and cataloging these patterns as to category. With a composer-specific lexicon and ATN grammar, the program is capable of randomly producing new works in the style of that composer. The program has produced works in the styles of composers such as Bach, Mozart, Brahms, Chopin, and Scott Joplin. Although none of these compositions could be mistaken for genuine works of the composers they are meant to emulate, the resemblance is at times quite striking. In particular, I found the works in the style of Brahms and Chopin surprisingly evocative. However, it is far from clear how the program produced these compositions. Composer Cope is clearly a convert to the joys of Lisp hacking, and he presents a simple version of his program virtually line by line, including many program traces. The more elaborate version of the program, which actually produced all the musical examples, is too large for such treatment. Instead, Cope contents himself with a vague, partial description of the program, leaving it quite unclear how the program actually produced all these interesting compositions. The most innovative aspect of this work is the notion that signatures, or common patterns, can be automatically extracted from a body of work. In the current implementation, the patterns are found by comparing simple sequences of intervals of a fixed length. Even with such simple comparisons, some striking regularities emerge, suggesting that this is a promising approach to the definition and analysis of musical style.--Daniel Hardt, University of Pennsylvania