The basic source material for musicology, and indeed for most musical performance (at least in classical music), exists primarily as notated scores. Historically, these have been produced over a period spanning nearly 500 years, and a great amount of scholarly and editorial effort has been applied during the past 200 years to the production of editions of this material. Vast quantities of musical scores are stored in the world’s libraries, yet only a tiny proportion of this huge resource is available in digital form, usually as image files. An even smaller amount can be found as encoded scores, that is, incorporating the structural and semantic knowledge about the music that a written or printed score imparts.
Since the advent of recording technology at the end of the 19th century, recorded (and, latterly, digitally-recorded) performances of classical music have become available in large quantity. The discipline of musicology is now beginning to recognize this audio legacy as an important resource in its own right. But, especially within the domain of classical music, the score is still of essential importance as both the generating source of the music (in the sense that performers generally play from scores), and as the most efficient and effective way of communicating and understanding the underlying structure and semantics of the music as opposed to the mere raw, undifferentiated signal data that is captured in a digital recording of a particular performance.
Recent technological advances have enabled the Humanities disciplines in general to benefit enormously from the possibilities offered by digital corpora (usually text-based and often supplemented by images or tabulated statistical data), but musicology has lagged behind in this process. Some of the pioneers in humanities computing, foreseeing the potential benefits, worked with musical materials from as early as the 1960s and music (in the form of audio recordings) was in the forefront of the digital revolution of the late 20th century. Yet there exist no large-scale digital corpora of music in the form of scores encoded in such a way that score-based and performance-based views of a musical work might be linked and analyzed together.
This project proposal aims to supply the methodological and technical foundation for the future building of such large-scale musical corpora which is our long-term goal. In the current phase we are preparing the ground by designing and implementing an infrastructure and carrying out some essential preliminary research and development.
Our approach offers major advances over current ad hoc methods. Building on the three-month Scoping and Feasibility study for which The Andrew W. Mellon Foundation provided funding, the work will proceed along three fronts: i) we will investigate and implement ways of improving current optical music recognition (OMR) technology and the means of encoding of OMR output in a form suitable for musical corpora; ii) we will develop and implement a novel representation scheme for musical works that will enable, for example, the alignment and cross-linking of score- and audio-based computer representations, thus allowing novel modes of presentation, processing and analysis of musical data, both within the single work and across collections of works; and iii) we will develop a musical corpus-building framework and workflow environment that will incorporate these innovations as they become available, together with other essential features (security, validation, access, etc.) in a robust, efficient and secure manner. At the same time we will consider some of the important aspects of a musicological methodology for working with musical corpora developed in such a system.
A musical corpus of modest size (up to 250 works), primarily intended to test the working of our systems, will be developed within the emerging framework. These will all be in the public domain, so as the test corpus grows it will have an additional "external" benefit in that it can be made freely available to scholars and other musicians and provide very useful test data ("ground truth") for those working in the audio domain, especially in music information retrieval (MIR).
Meeting Presentations
MeTAMuSE (Byrd/Knopke)
MeTAMuSE (Crawford)
Music Representation (Crawford)
