“Bridging disciplinary boundaries: A case study of the Music Information Retrieval Annotated Bibliography Project”

J. Stephen Downie Graduate School of Library and Information Science, University of Illinois at Urbana-Champaign jdownie@uiuc.edu

Background

Music information retrieval (MIR) research and development strives to afford the same level of access to the world's corpus of music as is afforded to text. Recently concluded international symposia devoted to MIR research and development bear witness to the promise of powerful, robust, large-scale, and flexible search and retrieval tools designed to be used by scholars and novices alike (ISMIR 2000; ISMIR 2001). Like many other humanities computing endeavours, it has attracted a multi-disciplinary group of researchers who are applying their domain-specific expertise to the wide range of challenges inherent in MIR development. MIR researchers come from library science, musicology, music theory, music psychology, computer science, digital and traditional librarianship, audio engineering, information science, information retrieval, publishing, media studies, broadcasting, law, and business, to name only a small handful of the disciplines represented. MIR research is all the stronger for the cross-pollination of disciplinary paradigms, research methods, and technological approaches. Effective communications across disciplinary boundaries, however, are impeding the coalescence of MIR research into a coherent discipline in its own right. The nascent MIR community is plagued with two forms of communications breakdown:

1. The scattering of the MIR literature caused by a lack of bibliographic control; and,
2. The confusion brought about by the use of discipline-specific assumptions, techniques, and language throughout its literature.

My paper reports upon the background, framework, goals and ongoing development of the Music Information Retrieval Annotated Bibliography Project (http://music-ir.org). It extends and augments the preliminary explication presented in Downie (2001). This project is being undertaken to specifically address and overcome the bibliographic control and communications issues plaguing the MIR research community. However, since most humanities computing projects also draw upon multi-disciplinary groups of researchers, I to hope provide and delineate a model that should assist others in facilitating efficient and effective communications among their particular research communities.

Problem Overview

MIR research has no disciplinary home. Because of this, researchers have been publishing their findings within the contexts of their own disciplines. Important MIR papers are thus scattered, for example, across the computer science, library science, musicology, humanities, and audio engineering literatures. There is no comprehensive indexing tool that successfully gathers up these papers. For example, significant musicology-based advances can be located through various music, arts and humanities indexes but not through the engineering and computer science indexes. Advances in audio engineering techniques are similarly absent from the music, arts and humanities indexes. Since researchers are generally unaware of the differences in scope of the various discipline-based indexes, they tend to focus upon those with which they are most familiar and thus overlook the contributions of those based in other disciplines. Unfamiliarity with the wide range of vocabularies used by the various disciplines further compounds the communication difficulties by making it problematic for MIR investigators to conduct thorough and comprehensive searches for MIR materials. Until these issues are addressed, MIR will never be in a position to fully realize the benefits that a multi-disciplinary research and development community offers, nor will it be able to develop into a discipline in its own right.

Solution

Our solution to the issues outlined above centers about the creation of a Web-based, two-level, collection of annotated bibliographies (Fig. 1). The first level, or core bibliography, will bring together those items identified as being germane to MIR as a nascent discipline. Thus, the core bibliography comprises only those papers dealing specifically with some aspect of the MIR problem, such as MIR system development, experimentation, and evaluation, etc. The second level, or periphery bibliographies, comprise a set of discipline-specific bibliographies. Each discipline-specific bibliography in the set will provide access to the discipline-specific background materials necessary for non-expert members of the other disciplines to comprehend and evaluate the papers from each participating discipline. For example, an audio engineering bibliography could be used by music librarians and others to understand the basics of signal processing (e.g., Fast Fourier Transforms, etc.). Another example would be a musicology bibliography that computer scientists could draw upon in an effort to understand the strengths and weaknesses of the various music encoding schemes, and so on. Thus, taken together, the two-levels of the MIR bibliography will provide:

1. The much needed bibliographic control to the MIR literature; and,
2. An important a mechanism for members of each discipline to comprehend the contributions of the other disciplines.

An important operating principle of the project is the use of non-proprietary formats and software. We are committed to the ideals of the Open Source Initiative (OSI, 2001) and the GNU General Public License (GNU Project, 2001) and thus intend to make our innovations freely available to others. In keeping with this commitment, we have chosen the Greenstone Digital Library Software (GSDL) package (New Zealand Digital Library Project, 2001), the Apache HTTP server (Apache Software Foundation, 2001), the PERL scripting language (PERL Mongers, 2001) and the Linux operating system (Linux Online, 2001) to create the basic technological foundation of the project. We have purchased copies of the commercial bibliographic software package, ProCite (ISI ResearchSoft, 2001) for initial, in-house, data-entry. ProCite also provides us with a representative instance of commercially available software that many end-users might utilize in manipulating the records they retrieve from our bibliography. At present, there are two central components of project undergoing development and alpha testing:

1. The bibliographic search and retrieval interface using the GSDL package; and,
2. The Web-based end-user data entry system.

For both of these, the goal is to create a system that will permit ongoing viability of the bibliography by minimizing-but not necessarily eliminating-the amount of human editorial intervention required. Item ^#1 issues being addressed include modifications to the basic GSDL system to permit the importation of specially structured bibliographic records and their subsequent access through a variety of field selection options. Item ^#2 is a CGI-based input system that guides end-users through the process of constructing well-structured bibliographic records through a series of step-by-step interactions and the on-the-fly generation of input forms specifically designed to provide the appropriate fields for the various types of bibliographic source materials (i.e., journal articles, conference papers, theses, etc.). Now that the general framework for the core bibliography has been laid, we are moving forward on the acquisition of the supplementary and explanatory materials. For these we are drawing upon the expert advice of those that have graciously signed on as advisors to the project. These advisors are not only lending their disciplinary expertise but are also affording us a very important multinational perspective on the potential uses and growth of the bibliography.

Acknowledgements

I would like to thank Dr. Suzanne Lodato for helping the project obtain its principal financial support from the Andrew W. Mellon Foundation. I would also like to thank Dr. Radha Nandkumar and the National Center for Supercomputing Applications Faculty Fellows Programme for their support. The hard work and insightful advice of Karen Medina, Joe Futrelle, Dr. David Dubin, and the other members of the Information Science Research Laboratory at the Graduate School of Library and Information Science, UIUC, is gratefully appreciated. Those members of the MIR research community who have volunteered to act as project advisors are also thanked.

Bibliography

Apache Software Foundation. Apache project. : , 2001.

J. Stephen Downie. “[Poster abstract] The music information retrieval annotated bibliography project, phase I.” Proceeding of the Second International Symposium on Music Information Retrieval, 15-17 October, 2001, Bloomington, IN. Bloomington, IN: Indiana University, 2001.

NU Project. Free software licenses. : , 2001.

ISI Researchsoft. ProCite. : , 2001.

ISMIR 2000. The International Symposium on Music Information Retrieval, 23-25 October, 2000, Plymouth, MA. : , 2000.

ISMIR 2001. The Second International Symposium on Music Information Retrieval, 15-17 October, 2001, Bloomington, IN. : , 2001.

Linux Online. The Linux homepage. : , 2001.

New Zealand Digital Library Project. About the Greenstone software. : , 2001.

Open Source Initiative. The open source definition. : , 2001.

Perl Mongers. Perl Mongers: The PERL advocacy people. : , 2001.