Stephen Downie: Large-scale music audio analysis: E-science or e-musicology?

Cross-posted from CIRMMT Events

The use of e-science technologies in the digital arts and digital humanities research domains represents a fast-growing area of scholarly activity. The "Structural Analysis of Large Amounts of Music Information" (SALAMI) project is an excellent example of the fruitful union of classic musicology and e-science. SALAMI is a multinational (i.e., McGill, Oxford, and Illinois) and multidisciplinary (i.e., music theory, library science, and computer science) digital humanities research collaboration. Exploiting 250,000 hours of compute time donated by the National Center for Supercomputing Applications (NCSA) at the University of Illinois, the SALAMI project is conducting the structural analysis of some 20,000 hours (i.e., roughly 2.3 years) of music audio. Thus, the SALAMI team is undertaking analyses of music structures at a scale that no individual human scholar could ever hope to undertake. This talk will also contextualize the SALAMI project within the broader frameworks of the ongoing “Networked Environment for Music Analysis” (NEMA) and the “Music Information Retrieval Evaluation eXchange” (MIREX) projects. The motivations, goals and developments of these three interrelated projects are presented to help illustrate the kinds of questions being explored by music informatics scholars and the roles that the e-science suite of tools, including high-performance computing, semantic web and Linked Data techniques, can play in answering those questions.