STATE-BASED SEQUENCING: DIRECTING THE EVOLUTION OF MUSIC

Traditional approaches to automatic music composition do not take into account the guided nature of music, in- stead augmenting existing material or generating scores based on provided seed parameters. Furthermore, these approaches often use a single algorithm to create a piece, where it is more natural to split the composition process into separate musical elements. Our new State-Based Se- quencer counters both of these limitations, using an agent- based collection of algorithms combined with a technique to provide motivation to the resultant music. In this paper, we describe the methodology behind this system, and pro- vide early results from an application of the framework. Our new distributed approach treats the existing algo- rithmic techniques as building blocks for the creation of a music composition system, where different algorithms can be plugged in for evaluation, while constraining the gen- eration through the use of a directing media. Rather than generating music with no prior information, a composer model(4) is used to provide a priori information for the algorithms, and a rich media ontology allows for the alter- ation of these parameters at pertinent points in the bound medium. This paper is split into three key parts, which follow the process of composing under the State-Based Sequencer. The rst describes the two representations that are required before composition can occur, namely the composer rep- resentation and the 'OntoMedia' ontological representa- tion. The OntoMedia representation details the compos- ing process itself and the agent framework that is used to provide the middleware for the distribution of composi- tion tasks. We then discuss the individual agents, showing how each is tailored towards a certain aspect of the mu- sic composition, and nally give some preliminary results from some of the implemented agents and discuss how the system will be developed further.