ÁñÁ«ÊÓƵ

Most previous books on computer music have focused on sound processing and synthesis techniques and most books on computer-assisted music analysis have dealt with music theoretical methods. Music Processing focuses on the methods of information and computer science that underlie music analysis and as such is a welcome contribution to the very new and rapidly growing discipline of cognitive musicology.

The book is well structured and features articles by some of the leading researchers in this area. The first section deals with the classic topics of computer music: music analysis, which is given a very good overview by Elaine Selfridge-Field, followed by articles on composition and sound processing.

The second section is concerned with methodologies, techniques and tools for music description and processing. This section starts with an article by Otto Laske on an epistemic approach to musicology, which is to my mind the most important article in the book in that it outlines the philosophical basis of a cognitive approach to music performance and understanding that transcends both empirical and humanistic science.

Laske argues that a humanistic science such as musicology should follow a methodology which permits it to show how the human mind works by investigating the way in which that mind is used in actual task performances. Musical action is dependent on large repertoires of concepts, schemas and strategies - mental constructs that musical actors rely on. Musical design programs can thus be seen as theories of musical action, and we can see the task of cognitive musicology as an action science in eliciting and elucidating such theories-in-action.

The essential idea is that musical theories-in-use can be seen to function as cognitive maps by which musicians design actions. The traditional division between musicological theory and musical practice has until now kept them hidden from view. These dimensions have now become more visible, thanks to theories-in-use enacted by living musicians, especially those using new technologies. By incorporating new technology, musicology can now concentrate more on the cognitive processes by which musicians produce works, rather than solely on the works themselves.

In the second article in this section, Marc Leman discusses the problem of describing and representing music. What is described in a traditional score? The music the composer had in mind? An externalised imagination or a translation in terms of a task instruction? How should musical information be represented in a computational model? As symbols interpreted by a closed system or at a sub-symbolic level that is integrally connected to the environment? These questions are discussed within a broader epistemological framework in a clear and concise way.

The article by editor Goffredo Haus and Antonio Rodriguez aims to understand and model the compositional processes of Ravel's Bolero so that suitable abstract tools can be defined for music description and processing. This is a good example of a traditional artificial intelligence approach to understanding cognition. In trying to understand the compositional processes at the structural level of music representation they attempt to decompose the work functionally into meaningful musical objects that can be described at various levels of abstraction within a hierarchical context of description. The aim is to understand the deep structure of the Bolero which lies in the synergy among various rules of development.

Antonio Camurri discusses the role of some knowledge representation paradigms and the role of time in AI formalisms, planning and action theories.

The final section includes four laboratory reports on the state of the art in computer music research at Carnegie-Mellon University, Stanford University's Center for Computer Research in Music and Acoustics, IRCAM in Paris and the Laboratorio di Informatica Musicale in Milan. These articles are all excellent in their description of the way in which cognitive music research has grown up in the main centres in the past 15 years, and are pointers to how new centres could develop in other universities.

This book is an excellent overview of the new field of musical informatics or music analysis by computer, a discipline which is still in its infancy and at a very exciting stage when looked at within the broader context of cognitive science. The presentation of the book is very well balanced in that it includes more philosophical articles, general overviews of the field and articles about specific implementations of methodology. The only criticism is that it has focused mainly on a traditional reductive approach to understanding cognition, which has been the predominant paradigm of the past 40 years. This approach assumes that deep knowledge can be abstracted by the modeller/programmer, encapsulated into objects and rules, and then be run on a computer to demonstrate intelligence. Now, rather than building closed systems whereby the programmer abstracts from the musical surface to rules and meta-rules in hierarchical programs, some researchers are now turning to the way in which higher-level abstractions emerge and evolve themselves without being constructed by "knowledge engineers".

Despite this omission this book is an excellent commentary on music description and processing and should be essential reading for anyone studying or researching computer applications in music, and anyone working on music-based multimedia.

Paul Hodgson is a musician, composer and a research fellow at the University of Sussex school of cognitive and computing science.