Vol. 35 Issue 2 Reviews

The Eleventh International Society for Music Information Retrieval Conference (ISMIR 2010)

Utrechts Conservatorium, Utrecht, The Netherlands, 9-13 August 2010.

Reviewed by KatieAnna Wolf
Cold Spring, Minnesota, USA

ISMIR 2010, the Eleventh International Society for Music Information Retrieval Conference, took place in Utrecht, the Netherlands, from 9-13 of August 2010.  It was jointly organized by Utrecht University, the Utrecht School of the Arts, the Meertens Institute, Philips Research, and the University of Oldenburg.  Since 2000, ISMIR has become an interdisciplinary forum dedicated to research on musical data, and brings together researchers in areas such as musicology, library and information science, cognitive science, computer science, and others.

I write this review from the perspective of a computer science student and a newcomer to the Music Information Retrieval (MIR) community. As a first-time ISMIR attendee, the conference offered me the opportunity to investigate the field and see the possibilities it had to offer. The goal of this review is to outline the events that transpired in Utrecht over the five days of the conference, and to present a fresh perspective on ISMIR, and on MIR in general.
As a student attending the conference, affordable accommodations were offered in student housing units. Being able to connect with other students the weekend before the conference began provided opportunities to not only socialize with others, new and experienced, in the community, but also to tour Utrecht and the surrounding cities in an enjoyable excursion prior to the start of the conference.

Before the official opening of ISMIR 2010, introductory tutorials were held on Monday, 9 August, for those interested in gaining background knowledge on some of the subfields within the MIR realm. One such session, entitled “A Tutorial on Pattern Discovery and Search Methods in Symbolic MIR,” presented by Ian Knopke (BBC) and Eric Nichols (Indiana University), focused on symbolic music representation and applications, and cognitive approaches to MIR, including an overview of the principles of music cognition and the use of symbolic models for various MIR tasks. As a part of the second round of tutorials Meinard Müller (Saarland University and MPI Informatik) and Anssi Klapuri (Queen Mary, University of London) presented “A Music-orientated Approach to Music Signal Processing,” which focused on explaining how music-specific aspects can be exploited for feature representation used in various MIR tasks.  The wide range of topics discussed included pitch and harmony, tempo and beat, timbre, and melody.  While I was familiar with most of these topics, the tutorial offered a well-rounded evaluation of areas that would be focused on throughout the conference. Overall, the tutorials gave an opportunity for those new to the field to gauge their knowledge and prepare them for the upcoming conference. The relaxed environment of the reception following the tutorials also offered an opportunity to get to know people in the community before the conference had formally begun.  

The theme for ISMIR 2010 focused on MIR research and applications that model the perception and cognition of music, that gave insight to the human musical experience and understanding, or that used creative innovations of MIR research.  To underscore this theme, Carol L. Krumhansl of Cornell University, well known for her research on tonal perception, opened the conference on Tuesday with her keynote speech entitled “Music and Cognition:  Links at Many Levels.” She explored the associations between the objective properties of music and the subjective experience of music. An example of such a link is her study on sensitivity to frequent patterns in sound events to the encoding and remembering of music and the generation of expectations. She showed a direct emotional response in fMRI (functional Magnetic Resonance Imaging) scans of the brain to music containing violations of those expectations. The links between objective properties and subjective experience, while present and important in the MIR community, seem to be becoming constantly more relevant in applied research as well.  Through the project Plink: “Thin Slices” of Music, she explored the identification of artist, title, and release date of short excerpts of popular music with results pointing towards a large capacity for details and knowledge of style and emotional content of music in long-term memory.

In connection to the perception of music, invited speaker Joris de Man, the composer for the Killzone video games, presented his talk on Thursday titled “Behind the Music of Killzone 2.” He offered an alternative perspective on how music is being incorporated into the technological world by discussing the process and tools used to create the music and set the mood within the computer game. In particular, he focused on the development of two types of music used in the game: linear music composed of live orchestral recordings used within the story cut scenes, and interactive music during play designed to react to what is happening in the game. 

Five poster sessions and four plenary sessions were held over the course of the conference, including two presentations from a new paper submission category titled State-of-the-Art Report (STAR). Intended to provide a review for the community on a variety of MIR subfields, the two STAR papers selected for publication out of seven submissions for ISMIR 2010 were titled “Music Emotion Recognition:  A State of the Art Review” (Youngmoo E. Kim et al.) and “Audio-Based Music Structure Analysis” (Jouni Paulus et al.). According to reports, of the 176 papers received for a double blind review (not including the seven STAR papers) 108 were selected for publication. Following a trend from previous years, most papers were presented as posters to allow for more intimate presentations and to facilitate more informal discussions, while also accounting for the vast range of topics. With numerous concurrent presentations on posters, the program, while dense, allowed attendees to target their exchanges. The papers selected for oral presentation during the plenary sessions were chosen only on the basis of representing the wide-range of interests, techniques and conclusions of ISMIR’s diverse multidisciplinary community.

As a unique element of the ISMIR conference every year, the results of the annual Music Information Retrieval Evaluation eXchange (MIREX) are released during a special poster session dedicated to the presentation of methods developed over the year for various MIR tasks. The exchange facilitates the evaluation and swapping of state-of-the-art MIR methods, which contributes to the dissemination of cutting edge techniques within the community so as to facilitate further advances in the field as a whole. Some of the tasks featured in MIREX 2010 included: audio classification tasks such as audio artist classification, various audio genre classifications, audio music mood classification, as well as audio onset detection, audio chord estimation, audio music similarity and retrieval, etc. 

Another notable segment of the conference was a morning dedicated to the future of MIR (fMIR). Two presentations kicked off the segment: the first on predicting the development of MIR research based on the parallels it has with natural language processing; and the second on how to address the challenges faced by MIR in order to make it a more versatile field. Douglas Eck then gave his talk on the future of MIR at Google with the idea that “music in the cloud” may be one direction the field is headed, in which the source of one’s music no longer sits on one’s personal device but on a public domain such as YouTube. Another point he touched upon was that the future may bring a closing of the link between listening and making music, which may expand MIR to another direction. An industrial panel on fMIR then followed, moderated by Rebecca Fiebrink (Princeton); the panelists consisted of Douglas Eck (Google), Greg Mead (Musicmetric), Martin Roth (RjDj), and Ricardo Tarrasch (Meemix). The group discussed topics involving how music technology would be changing over the next few years. With the increase in the computing power at people’s fingertips through their cell phones, the panel expanded on the possibilities for more user-defined personalization of listening capabilities. Another area touched upon is the possibility of the shifting of roles of the people within the music world, as listeners become composers via music generation, or produce music based on the environment. It was encouraging to know that leaders in the field are looking ahead to what it is becoming and how it is evolving.              

Another aspect of the conference involved celebrating the enjoyment of music through several recitals held at the close of each day, after the posters and presentations. A reception commemorating the first day of the conference was sponsored by the City of Utrecht in the garden of the “Academie Gebouw” of Utrecht University during which Arie Abbenes performed on the Hemony Carillon of the Dom Tower. The performance comprised of a combination of pieces by W. A. Mozart, in honor of the unofficial ISMIR theme, “Ah vous dirai-je, Maman,” and by Jacob van Eijck, as well as songs of The Beatles. The carillon recital demonstrated unequivocally how music surrounds daily living as the bells rang out for the entire city to hear. Wednesday’s recital, presented by the Utrecht School of Music and Technology (USMuT), one of the schools of the Utrecht School of the Arts (HKU), featured Sonsoles Alonso, an emerging pianist specializing in contemporary music and live electronics, and three works by composers combining technology and sound in innovative ways. The recital encapsulated the various ways in which music and electronics can be combined together to contribute to an emerging genre of performance for the future.
The conference banquet, held Thursday night, was hosted at, and included guided tours through, the National Museum “From Musical Clock to Street Organ.” The attendees were treated to demonstrations of the various mechanical instruments, ranging from carillon clocks and music boxes to street and dance organs. It was a reminder of how music has evolved over the years, and how those at the conference are contributing to that development through their research.  Concluding the dinner was the announcement of the best student paper, awarded to Hussein Hirjee and Daniel G. Brown for their paper “Solving Misheard Lyric Search Queries Using a Probabilistic Model of Speech Sounds,” and the best overall paper, awarded to Ron J. Weiss and Juan Pablo Bello for their paper “Identifying Repeated Patterns in Music Using Sparse Convolutive Non-negative Matrix Factorization.” 

As a fitting conclusion to the conference on the morning of Friday, 13 August,  the late-breaking/demo session offered the latest developments in the MIR field, including hot-off-the-press research conducted in the recent months leading up to the conference. The session presented a tantalizing glimpse of what one could expect at the upcoming ISMIR 2011. After the final plenary session, the closing remarks wrapped up the conference with many thanks to the staff and an announcement of next year’s conference to be hosted in Miami, Florida, USA, during the last week of October 2011 with the theme “Music, Anytime, Anywhere.”

The welcoming environment of ISMIR 2010 and the amount of exchanging and developing of ideas within the field were convincing factors in determining the success of the conference. With such an international following and a diverse collection of subfields, the MIR community stays in continuous communication with a total of 1,346 subscribers to their online mailing list (subscription details at www.ismir.net).  For those interested, it offers updates in the field and research discussions, as well as announcements and information on upcoming conferences.