Page  241 ï~~THE USEFULNESS OF QUA OF MUSICAL PE Gerhard WIDMER (gerhard@ai-vie.uucl Austrian Research Institute for Artificial Department of Medical Cybernetics and University of Vienna Austria ABSTRACT: The paper discusses the ic qualitative theories of (tonal) music as a puter systems. In particular, it concentra theories for programs that are to learn ab' tive theories are a very natural formalisr knowledge about music, that such theorie process in a learning system, and finall? interectino insiht frrnm Pvnrimnntal nnr I

Page  242 ï~~to musicians in intelligible, common-sense' t learn new musical concepts effectively. Obvi the a specific knowledge. order to keep heavily relies on methods and techniques fr idea is to give a learning program all the k readily available and can easily be formulat itself. For this learning scenario not to look knowledge given to the system to be some 'basic' knowledge the kind of knowledge Here interesting question arises: is then regarded as a psychologically and musicologi, indeed such basic musical 'knowledge' (ever sum total of what might be called 'habits c tonal music perceive and organize music in tl 1983) from, It is the kind of 'knowledge' that orn years of exposure to tonal music and rnontc c ir'h ac irktk~rar mi icirsal n'acca o cn

Page  243 ï~~The perception model for harmonized melodi of an effect. Effects are sensations evoked ir event. Such effects can arise on different Zev sive musical events that are very different in trast; harmonic progressions may be felt to % abstract level, contrast, diversity, and buildup an overall feeling of interestingness of a pass on various hierachical levels of abstraction, ai ences between effects (for instance, whether weakens the perceived prominence of some For instance, the following qualitative state smoothness of the transition between adjacen musical passage: QUAL-POS-PROP ( smoothness-of-trar ("perceived smoothness of transition from a positively proportional to perceived cohe

Page  244 ï~~Apart from experiments of theoretical interest applications in practical musical systems. For i model (or of a more complex model that coy could be exploited in intelligent musical assistar music education, etc. In all these scenarios, a c able the system to evaluate given musical piec their plausible quality, and explain its decisions gible terms. Possible benefits of such projects fc Turning to more general issues, it seems clear th ogy of music can reap some important benefits f ries of musical perception and organization. T ments combining such qualitative theories and light on the adequacy of the underlying musica musical style are easily learnable on the basis c natural connection between the assumptions abc