Optical Music Recognition for Plainchant

Project Overview

Plainchant is a large collection of monophonic songs, which exist as one of the oldest notated music in the world. The music has been part of Christian liturgy since the Middle Ages and has formed the basis for much Western music that followed. Canada is positioned to become a world leader for studying plainchant. Combining information from the University of Western Ontario's CANTUS database, one of the world's foremost databases of metadata about plainchant, with our expertise as two of Canada's leading scholars in optical music recognition and early music, we are developing a network and database infrastructure and software tools that will be able to automatically extract the musical and textual content from digitised plainchant manuscripts and to index that content in an internationally-accessible distributed database. This recognition system will free libraries from the prohibitive cost of expert human labour needed to transcribe these specialised manuscripts manually and will allow performers and coaches, such as those at the Gregorian Institute of Canada, to expand the repertoire and improve the historical fidelity of their recordings and performances.

Plainchant manuscripts, copied by hand from the ninth through the sixteenth centuries, constitute some of the richest source material available for musicological scholarship and historically informed music performance. Unfortunately, because these sources are spread throughout libraries and religious institutions worldwide, it is almost impossible for scholars to form a complete picture of the development of plainchant regionally and chronologically. Facsimiles are sometimes available for local libraries to purchase, but they tend to be exceedingly expensive and, like other paper documents, cannot be searched or indexed against other sources; microfilms are more affordable and more often available but suffer the same difficulty of searching and indexing. Many chants appear in multiple variations across numerous manuscripts, and thus searchable, cross-referenced digital libraries of musical content will be especially valuable for these documents. The main goal of our project is to facilitate the creation of such a digital library by introducing optical music recognition (OMR) tools to reduce the need for human transcription, much as optical character recognition (OCR) has done for text documents.

As part of our most recent SSHRC Standard grant, we have developed an OMR system that attains a 97 percent recognition rate for printed symbols from later Western music notation, and with our current SSHRC Image, Text, Sound, and Technology grant, we are developing a system capable of recognising printed lyrics. Our new project will begin by extending this system to recognise printed square-note neumes, which became the most common system of hand-written notation for plainchant around 1200. Although this notational system is centuries old, it continues to be used today when publishing compilations of liturgical plainchant, e.g., the Liber Usualis, a common reference book for practising church musicians and plainchant scholars. After determining the best strategy for recognising printed neumes, we will extend the work to hand-written sources, focusing on three distinct notational styles: hand-written square-note neumes, analogous to printed chant notation; Aquitanian neumes, an older, staffless notational style in which individual neumes are composed of separated strokes and points; and Gothic neumes, variations on the square-note neume shapes that we expect to be more difficult to recognise.

Interdisciplinary in its very nature, this project will bring together researchers at all states of their careers, from both musicology and computer engineering. In order to provide the best possible training environment for students and in order to strengthen the connection between these two disciplines, the project will include regular presentations of intermediate results to all communities involved. When complete, it will have produced infrastructure and tools ready for musicians and others to generate and share digital libraries comprising centuries of plainchant and will open a wide range of new avenues for musicological research.

People

Researchers

Doctoral Students

Masters & Undergraduate