Music score interpretation using a multilevel knowledge base

The problem of interpretation of musical scores is a typical artificial intelligence problem. In fact it is poorly defined and not explicitly formalizable. Moreover it requires wide knowledge domains: cultural, stylistic, of instrumental technique, of sensitiveness etc. We focused our interest on the interpretation of computer music. An approach to build a rule system based on many sorted logic is presented. This approach allows to work with partial knowledge and to reply to a general query. This ability is obtained by a hierarchical organization of the knowledge domain. The organization of the knowledge base and the methodology employed to formu- late the rules are described. Two different knowledge bases can be distinguished. The musical knowledge base codifies the experience and fantasy of the interpreter (e.g. conductor) and the sonologic one collects the competence of the performer (e.g. orchestral musician). Finally a prototypical system based on this approach is described.