Cluster on Computation in Music Guest Editor: Elaine Chew ( echew (at) usc.edu ), University of Southern California Associate Editors: Roger B. Dannenberg ( rbd (at) cs.cmu.edu ), Carnegie Mellon University Joel Sokol ( jsokol (at) isye.gatech.edu ), Georgia Institute of Technology Mark Steedman ( steedman (at) inf.ed.ac.uk ), University of Edinburgh

CALL FOR PAPERS ( to PDF or plain text )

The issues in computational modeling of music have risen to the forefront due to the entertainment sector's interests in music recommendation (personalized systems for online retrieval of music) and performance rendering systems (affecting the movie scoring and music industries). The need for new algorithms and models better to understand, compute, and evaluate the structure and effects of music is made more urgent by the rapid proliferation of digital music. Many problems in computer-based methods for content analysis and feature extraction, similarity assessment and classification, and generative methods for creating music can be and have been solved using core OR approaches.

We invite papers that highlight the use of computational operations-research techniques in analysis and generation of music. The broad topic areas include, but are not limited to, computational modeling of music perception and cognition, music information retrieval, computer-assisted composition, and interactive music systems. The goal of the cluster is to introduce computational research in music to the OR community at large. As such, the papers should not only feature the contributions of the author(s) to the domain but also provide a clear and expository review of key advances in the specialty area and identify or define some concrete and open problems.

By operations-research techniques, we mean mathematical, statistical, and computational methodologies defined and documented in the OR literature, such as mathematical programming (linear, nonlinear, integer), network models, and other deterministic techniques; dynamic programming, Markov chains, queuing theory, and other stochastic processes; and more recent inventions such as genetic and evolutionary algorithms, simulation, neural networks, approximation algorithms, and domain-specific heuristic methods.

Some sample (and overlapping) topics include the following:

• Music representation and comparison. Includes various ways to represent music and the reliability and accuracy issues associated with each representation, and ways to detect and assess similarity for classification and categorization. Methods in use include clustering and mathematical-programming techniques and combinatorial approaches. Algorithmic efficiency and scalability is an important issue with rapidly expanding digital databases.

• Tonal structure and function. Topics include recognition and prediction of pitch structures and functions, such as key-finding, tonal partitioning, and chord analysis. Existing methods include linear and dynamic-programming techniques, linguistic approaches, and neural network models. Determining accurate solutions to the problems is challenging due to uncertainty, for example, incomplete or extraneous information.

• Time structure detection and prediction. An essential part of music is the beat and meter. Recognition problems related to these time structures include beat tracking, meter induction, and rhythm analysis. Methods include tree-based approaches, multi-hypothesis evaluations, rule-based methods, and neural networks.

• Computer-assisted composition and improvisation. Stochastic methods for generating viable solutions in a large solution space have long been employed in music composition. Other techniques include network models, constraint programming approaches, genetic and evolutionary algorithms, and Markov-chain models.

• Performance-rendering sytems. Performance is the manipulation of a variety of factors such as timing of beats, dynamics, and articulation. An interpretation is the result of one particular sequence of inter-related decisions. Existing methodology for generating expressive performances include case-based reasoning, neural networks, and probabilistic approaches.

The INFORMS Journal on Computing is a publication of the Institute for Operations Research and the Management Sciences (INFORMS). The JoC publishes results in the intersection of operations research and computer science.

You can submit electronically a postscript or pdf file to the Editor, or to any of the Associate Editors with Cc: echew (at) usc.edu. Direct your questions to:

Please post and distribute this announcement freely. Last update: January 15, 2004.