A Structured Framework for Representing Time in a Generative Composing System

The representation of music structures is, from Musicology to Artificial Intelligence, a widely known research focus. It entails several generic Knowledge Representation problems like structured knowledge representation, time representation and causality. In this paper, we focus the problem of representing and reasoning about time in the framework of a structured music representation approach, intended to support the development of a Case-Based generative composing system. The basic idea of this system is to use Music Analysis as foundation for a generative process of composition, providing a structured and constrained way of composing novel pieces, although keeping the essential traits of the composer’s style. We argue that, for a problem of this kind, it is important to provide both the means to represent the structure of music and its antecedent-consequent relations, and to represent and reason about time relations, even when dealing with incomplete information, as when incrementally constructing a solution from the adaptation of illrelated pieces. We propose a solution which combines a tree-like representation with a pseudo-dating scheme to provide an efficient and expressive means to deal with the problem. We also show that it provides a convenient framework for representing knowledge in other domains structured in a similar way.