Composing Music by Composing Rules: Design and Usage of a Generic Music Constraint System

This research presents the design, usage, and evaluation of a highly generic music constraint system called Strasheela. Strasheela simplifies the definition of musical constraint satisfaction problems (CSP) by predefining building blocks required for such problems. At the same time, Strasheela preserves a high degree of generality and is reasonably efficient. Strasheela is highly generic, because it is highly programmable. In particular, three fundamental components are more programmable in Strasheela compared with existing music constraint systems: the music representation, the rule application mechanism, and the search process. Strasheela features an expressive symbolic music representation. This representation supports explicitly storing score information by sets of score objects, their attributes, and their hierarchic nesting. Any information available in the score is accessible from any object in the score and can be used to obtain derived information. The representation is complemented by the notion of variables: score information can be unknown and such information can be constrained. This research proposes a rule formalism which combines convenience and full user control to express which score variable sets are constrained by a given rule. A rule is a first-class function. A rule application mechanism is a higher-order function. A rule application mechanism traverses the score in order to apply a given rule to variable sets. This text presents rule application mechanisms suitable for a large set of musical CSPs and reproduces important mechanisms of existing systems. Strasheela is founded on a constraint programming model which makes the search process programmable at a high-level. The Strasheela user can optimise the search for a particular constraint satisfaction problem by programming a distribution strategy (a dynamic variable and value ordering) independent of the problem definition. Special distribution strategies for efficiently solving various musical CSPs – including complex polyphonic problems – are presented.