COMPOzE - Intention-based Music Composition through CO t Pn, Martin Henz Stefan Lauer Detlev Zimmermann Programming Computer Science Graduate School for Systems Lab Department Cognitive Science University of Saarland Im Stadtwald D-66041 Saarbriicken, Germany E-mail: {henz,lauer, detlev}@cs . uni - sb. de Abstract The goal of this work is to derive four-voice music pieces from given musical plans, which describe the har- monic $ow and the intentions of a desired composition. We developed the experimentation platfomz COMPOzE for intention-based composition. COMPOzE is based on con- straint programming overJeinite domains of integers. We ar- gue that constraint programming provides a suitable tech- nology for this task and that the libraries and tools avaib- able for the constraint programming system Oz efectively support the implementation of COMPO&. This work links the research areas of automatic music composition on one hand adJeinite domain constraintpro- gramming on the other, and contributesthe tool COMPOzE, which practically demonstrates the potential of constraint programming to open up new areas of application for auto- matic music composition. 1 Introduction The aim of this project is to build a system for the auto- matic composition of music. Music experts are often skeptical about music which is autonomously composed by computers. We share this skepticism and therefore chose-instead of completely au- tonomous composition-the task of composition to accom- pany multimedia presentations as our application domain. In this context, music serves as acoustic background, and supports the intentions of the presentation using the appro- priate musical effects. The system developed in the project consists of two main modules: the arrangement system AARON [U] and the compositionsystem COMPOzE. AARON derives a musical 0-8186-7686496 $05.00 0 1996 IEEE plan from the intentional and metric structure of a given pre- sentation by proceeding in two steps. In the first step, it gen- erates a vector of musical parameters describing in musical terms how the given intentions ought to be realized. In the second step, it generates from these musical parameters an harmonic progression. The harmonic progression fixes the metric, rhythmic and harmonic svucme, but leaves open the harmonic elaboration and the melody. COMPOZE derives concrete audible music (i.e. MIDI data) from a musical plan and an harmonic progression. In this presentation, we focus on the COMPOzE subsystem. COMPOzE produces a progression of four voice chords (soprano, alto, tenor, bass), which implements the musical plan generated by AARON and is in accordance with stan- dard musical laws. To accomodate different musical tastes and to allow for easy tuning of the system, we want to open up the composition process to the user by giving her max- imal flexibility in the choice of the musical laws. COM- POzE’s graphical user interface allows the user to choose musical laws by direct manipulation. The composition pro- cess is visualized including its solutions. The user can listen to and compare the solutions by mouse click. It turns out that this task of open intention-basedcompo- sition can be elegantly described as a constraint satisfaction problem and efficiently implemented using the constraint programming (CP) system Oz. The main contribution of COMPOzE is to demonstrate that CP in general and Oz in particular provide an adequate computationalframework for open intention-based music composition. 2 Musical Plan The given musical plan consists on the one hand of an harmonic progression. An harmonic progression is a se- quence of harmonic functions. As an example, consider: 118