jmj@wpi.WPI.EDU (John M. Joy) (02/05/91)
I am currently pursuing work on a computer science MS thesis on the feasibility of creating an automated analysis system a la [Lerdahl & Jackendoff, "A Generative Theory of Tonal Music"]. Does anyone know of any current (or past or future) work in the same or similar area? I have already perused Computer Music Journal and Computers & the Humanities, as well as numerous texts relating to analysis and generic uses of computers in music research (BiBTeX available on request). Incidentally, yes, I DO know that the theory (as given in [L&J83]) does not constitute an effective procedure: 1) presence of informally defined concepts (e.g. parallelism); 2) lack of conflict resolution between preference rules; 3) potential "well-formed" analyses for a given piece represent an infinite search space. 1) will be treated as a peripheral issue (there is enough research on cluster analysis and musical applications and the like (such as [Mongeau&Sankoff, CHum 24:161-175]) to provide an algorithm for computing something "reasonably close" to [L&J83]'s intended meaning of "parallelism"). 2) will be handled by pursuit of means of "weighing" (functionally?!? I doubt that the preference rules are context free). As for 3), this frees me from finding the optimal analysis, a satisfactory one is good enough (i.e. I must define "satisfactory"). Any information or leads (not noted above) would be appreciated. Please respond to me at: jmj@maxine.wpi.edu or: jmj@wpi.wpi.edu Thanks. JMJ P.S. The thesis is NOT intended to deal with the suitability or usefulness of the [L&J83] analysis, so please no flames!