Preliminary report on the design of a constraint-based musical planner

Holland, Simon (1991). Preliminary report on the design of a constraint-based musical planner. Department of Computing Science, King's College.


This work described in this paper forms part of a wider research project, described in Holland (1989), to find ways of using artificial intelligence methods to encourage and facilitate music composition by musical novices. This paper focusses on the key component of a knowledge-based tutoring system under development to help novices learn to compose and analyse musically 'sensible' chord sequences. This key component is a constraint-based musical planner dubbed 'PLANC'. The musical planner (together with its set of musical 'plans') can be used to construct and analyse chord sequences in terms of musical strategies that can be understood and made use of by complete musical novices. PLANC can generate a class of musically 'interesting' chord sequences that include the chord sequences of many well known existing pieces of music, as well as generating a large space of new 'interesting' sequences. The design of the planner draws on a characterisation of creativity due to Johnson-Laird (1988). The planner is psychologically plausible , though not intended as a detailed cognitive model. An overview of the structure of PLANC is presented, and its suitability for use in a tutoring system is considered. The design of the planner is criticised. Each of the main components of PLANC is analysed: plan variables, constraints, value generators and methods. Much of the 'knowledge' used in PLANC consists of informal musical knowledge: three appendices analyse the different kinds of informal knowledge used. The applicability and value of similar constraint-based mechanisms in intelligent tutors in a wide range of other open-ended domains is considered

Viewing alternatives

Download history

Item Actions