A Prototyping and Evaluation Environment for Music Digital Library Research
This demonstration introduces the M2K (Music-to-Knowledge) music digital library (MDL) and music information retrieval (MIR) prototyping and evaluation environment. M2K is being developed as part of the International Music Information Retrieval Systems Evaluation Laboratory project (IMIRSEL). The goal of the IMIRSEL project is the creation of standardized tools and datasets for the scientific evaluation of MDL/MIR systems. M2K played a central role in the Music Information Retrieval Evaluation eXchange (MIREX) contest held over the summer of 2005. MIREX organizers used M2K to run the formal evaluation tests on the submissions that tackled such MDL tasks as artist identification, genre classification, etc. The MIREX participants met in September, 2005 at the 6th International Conference on Music Information Retrieval (ISMIR 2005) in London, UK.
The computational backbone of M2K is the Data-to-Knowledge (D2K) machine learning environment developed by the Automated Learning Group (ALG) of the National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign (UIUC). The D2K framework allows developers to "wire" together computational modules into programs called itineraries (Figure 1) which represent dataflow between modules. These itineraries can then nested within other itineraries and used as modules, allowing for the development of itineraries with arbitrary complexity (Figure 2).
M2K is being released as an open source distribution to encourage the development of MDL-specific modules. Information about M2K is available at <http://music-ir.org/evaluation/m2k>.
Project funders: The Andrew W. Mellon Foundation, and the National Science Foundation (IIS-0340597 and IIS-0327371). K. West (University of East Anglia) and P. Lamere (Sun Labs) are thanked for programming contributions. D. Tcheng, J Futrelle, M. Welge, L. Auvil and D. Cai, all of ALG, are thanked for their assistance.
© Copyright 2005 J. Stephen Downie, Andreas F. Ehmann, and Xiao Hu