David Huron
Computers in Music Research Conference, 1988.
The problem of computer-based music representation is considered through an analysis of signifiers, signifieds, and denotative mappings. Binary, alphanumeric, graphic, and auditory signifiers are discussed, and seven classes of denotative mappings distinguished. Various design principles for music representation are developed. Representational protocols are advocated over representational schemes as protocols permit individual users to tailor their own "schemes" specifically to suit the type of task in which they are engaged, and yet produce representations which remain compatible in principle with other forms of representations.
The 1960s and early '70s saw an initial flurry of activity pertaining to the problem of computer-based music representation. The debates of that era have remained dormant until the recent creation of the ANSI working group X3V1.8M on a possible standard for computer-based music representation. In this paper, it is argued that the problem of music representation holds a central position in computer-based musical activities -- both with respect to research as well as with respect to creative endeavors.
The paper begins with a critical review of the design philosophy which led to the creation of the DARMS representation language by Stephen Bauer-Mengelberg and Ray Erickson. It is argued that DARMS, ALMA, MUSTRAN, ANON, CLML, and similar languages suffer from four specific problems: high context-dependency, structural convolution, inflexibility regarding types and levels of detail, and (in the case of DARMS) the mistaken belief in a non-interpretative representation. Predictably, the ensuing complexity has discouraged the writing of analysis software which works directly on the representation data. Attempts to build more amenable data structures and databases would be less pressing if the representation scheme itself was better structured. The most potent testimony to the shortcomings of DARMS-like representational schemes is the great number of scholar-specific dialects which have arisen in an attempt to reduce the complexity of analysis software and to maintain some dgree of productivity in the light of certain goals.
Starting from different design premises, an alternative representation approach is developed. The notion of a representation scheme (as in an icon-ASCII mapping) is replaced by the idea of a representational protocol. Individual users are encouraged to design their own "schemes" within the guidelines of the protocols, specifically to suit the type of task in which they are engaged. Thus a schlar interested in handwriting analysis of manuscripts may proceed with computer-based methods undeterred by the fact that existing schemes discard crucial detail. Conversely, a research interested in Beethoven's orchestration can proceed without incurring the laborious penalty of encoding all of Beethoven's orchestral works to a level of prescribed detail incommensurate with the envisioned form of analysis. A representational protocol (as opposed to a representational scheme) permits a musician or scholar to focus on those details (or generalities) germane to the task at hand, whilst ensuring that different representation "schemes" for a given work are mutually compatible. Hence it is possible to conceptualise an integrated network of music representations from the discrete sound function, sound synthesis `note records', and MIDI, to the traditional score, analytic representations, and compositional `meta-scores'. What we call "musical knowledge" may thus be conceived as the means by which one form of representation is translated into another form.
Abstract from subsequent 1989 publication.