STATISTICAL RULES IN CONSTRAINT-BASED PROGRAMMING

In this paper we introduce a system that first generates stati stical analysis data from a musical score. The results are then translated automatically to constraint rules that in turn can be used in com- bination with ordinary rules to generate scores that have si milar statistical distributions than the original. Statistical analysis rules are formalized using our special rule syntax where our focus will be in the pattern-matching part of the rules. The pattern-matching part has two important tasks in our paper: first, it is used to e xtract various musical entities from the score, such as melodic, harmonic and voice-leading formations; second, it is used also to generate statistical rules which will be used in the re-synthesis par t of our system. We first introduce the rule syntax. After this we disc uss a practical case study where we analyze a melodic line. Finally we generate out of this material statistical rules which are used to produce new scores.