GSLIS Professor and Associate Dean for Research J. Stephen Downie will deliver a keynote address at the Brazilian Symposium on Computer Music (SBCM), which will be held October 31 to November 2 at the Escola de Música de Brasília in Brasília, Brazil. The invited talk, “Using MIREX to Examine the Evolution of Music Information Retrieval Research,” will be given on November 2:
Abstract: The "Music Information Retrieval Evaluation eXchange" (MIREX) is the annual cycle of events wherein music information retrieval (MIR) researchers come together to investigate how well their innovative MIR algorithms perform. Since its launch in 2005, MIREX has been directed by J. Stephen Downie at his International Music Information Retrieval Systems Evaluation Laboratory (IMIRSEL), Graduate School of Library and Information Science (GSLIS), University of Illinois at Urbana-Champaign. MIREX has played a pivotal role in the growth and success of the Music Information Retrieval (MIR) research community. MIREX has evaluated almost 1600 algorithms across a wide range of MIR task categories. This talk will introduce the remarkable diversity and evolution of MIR research problems by examining the history and development of MIREX. We will show how MIREX has been able to track the remarkable progress that has been made, for example, in such areas as audio melodic extraction, audio and symbolic genre classification, audio chord detection, audio mood tagging, audio and symbolic music similarity, and audio structural segmentation. See the MIREX wiki for more information about MIREX.
In addition to the invited talk, Downie will lead a tutorial titled, “A Demonstration of Audio Music Discovery, Classification and Analytic Tools”:
Abstract: Interest in music information retrieval (MIR) research has grown over recent years. Exciting advances in the music audio domain have garnered special attention. As this interest has grown, new audio-based MIR tools and dataset opportunities have been made available. This demonstration will survey a sample of readily available—mostly open source—music audio tools and datasets that can be used by newcomers to MIR explore a wide range of music audio discovery, classification and analytic options. The workshop will introduce possible data and metadata resources. Music audio feature selection and extraction tools will be demonstrated. Sample music classification and analytic experiments will be run. Select higher-level music digital library prototypes will be presented to illustrate of the new music audio discovery and exploration functions that MIR research are making possible.