On July 12th and 13th 2012 we organized the 2nd CompMusic Workshop in Istanbul, at Bahçeşehir Üniversitesi. Right before the workshop we also had three days of meetings with all the project team. This was the first time we had all the members of CompMusic together. The location was idilic, right on the Bosphorus, with a great view of the strait and of the city. Both the pre-workshop meetings and the actual workshop were very successful, academically and socially. Thanks to the great organization by the team of Bariş Bozkurt everything went very smoothly and the feed back by the participants has been very positive.
This second workshop coincided with the starting of the second year of the project, an ideal time for planing the remaining four years while being able to present some initial results to prove the feasibility of the project goals. We are now four research teams and we are working on three (Hindustani, Carnatic and Turkish-makam) of the five chosen music cultures. It is still too early to evaluate project results but it was great to see the work already done and the increasing clarity of the project focus.
The topics and research approaches than can be covered in CompMusic are too many and we can only focus on a few. The presentations by the project members showed this focus and it was good to have presentations by external researchers that covered other relevant topics. In the project we are mainly focusing on: intonation characterization, melody and rhythm description, community characterization, and music browsing and discovering methods.
On the pre-workshop meetings we had overview talks by project members, giving space for informal discussions and exchanges of methodological approaches among the project members. The agenda of these meetings was:
- Tools (July 9th): CompMusic Browser and computational infrastructure at UPF (M. Sordo); MusicBrainz (A. Porter); Freesound (F. Font); Machine learning methods for music analysis (H. Murthy).
- Pitch and rhythm analysis (July 10th): Intro to pitch analysis (P. Rao); Comparison of pitch analysis algorithms (S. Gulati); Pitch analysis algorithms details (P. Rao, B. Bozkurt, S. Gulati); Rhythm issues (A. Holzapel, A. Srinivasamurthy); Turkish rhythm analysis [slides, video] (A. Holzapel, Fahrettin Yarkın); Indian rhythm analysis [video] (A. Srinivasamurthy).
- Music discussions (July 11th): Open discussion on Turkish music with Okan Murat Ozturk; Open discussion on Carnatic and Hindustani music with T M Krishna and Suvarnalata Rao.
The actual workshop included 28 talks, most of them on research carried out within CompMusic but we also had presentations by external researchers working on topics of relevance to the project. For the opening of the workshop I gave a talk on "Opportunities for a Cultural Specific Approach in Computational Description of Music" in which I motivated and summarized the main research areas that we are working on. This was followed by three talks covering the musical traditions being studied. Suvarnalata Rao talked on "Culture Specific Music Information Processing: A Perspective from Hindustani Music", T M Krishna on "Carnatic Music: Svara, Gamaka, Motif and Raga Identity" and Okan Murat Ozturk on "A Semiotic Approach to the Analysis of Makam Melodies: The Beginning Sections of Melodies as "Makam Indexes"". The three musicians have become collaborators of CompMusic and their presentations showed how successful this collaboration is. The remaining workshop presentations dealt with specific technical topics that I list here.
Tonic detection and Intonation characterization in Indian music:
- A. Bellur, V. Ishwar and H. Murthy. "A Knowledge Based Signal Processing Approach to Tonic Identification in Indian Classical Music".
- S. Gulati, J. Salamon and X. Serra. "Tonic Identification System for Hindustani and Carnatic Music".
- G. K. Koduri, J. Serrà and X. Serra. "Computational Analysis of Intonation in Indian Art Music".
In Indian music the tonic functions as the reference pitch from which all the harmonic and melodic relationships are stablished during a performance. Its automatic detection is a necessary step for analyzing many musical aspects and thus it was the first problem that we wanted to solve. I consider the tonic detection results obtained and reported in these articles sufficient for the purposes of the project. A natural next step is the characterization of the intonation used in a performance. This is a not solved problem but the initial results reported at the workshop are quite promising. The proposed representation of intonation in raga music can be a valuable element used to describe various musical aspects.
- B. Bozkurt. "Features for Analysis of Makam Music".
- J. Ch. Ross and P. Rao. "Melodic Phrase Segmentation from Audio in Hindustani Classical Music".
- A. Vidwans, K. K. Ganguli and P. Rao. "Detecting Indian Classical Vocal Styles from Melodic Contours".
- V. Ishwar, A. Bellur and H. Murthy. "Motivic Analysis and its Relevance to rAga Identification in Carnatic Music".
- E. Ünal, B. Bozkurt and M. K. Karaosmanoğlu. "Incorporating Features of Distribution and Progression for Automatic Makam Classification".
- E. Özek. "The Concept of çeşni in Turkish Music and the Analysis of Performance-Theory Differences".
- T. H. Özaslan, X. Serra and J. Ll. Arcos. "Signal Analysis of Ney Performances".
- S. Şentürk, A. Holzapfel and X. Serra. "An Approach for Linking Score and Audio Recordings in Makam Music in Turkey".
- M. Subramanian. "Generating Computer Music from Skeletal Notation for Carnatic Music Compositions".
- S. K. Subramanian, L. Wyse and K. McGee. "A Two-Component Representation for Modeling Gamakas of Carnatic Music".
There were many talks related to the characterization of melodies, both in Indian and Turkish musics. An advantage of Turkish music is the existence of scores, making it very valuable for the study of some melodic aspects. For using the audio signals an important first step is the extraction of the pitch contour of the lead instrument, which requires a multipitch methodologies. Both makam and raga music have particular ways in which melodies are constructed and thus they require specific approximations in order to characterize them. Most of the presentations reported preliminary work and there is a long way to go before we have proper parameterizations with which to identify musically meaningful melodic elements. Two of the papers by external researcher focused on the synthesis of melodies, which is an interesting aspect not covered in the project.
- A. Srinivasamurthy, S. Subramanian, G. Tronel, P. Chordia. "A Beat Tracking Approach to Complete Description of Rhythm in Indian Classical Music".
- A. Holzapfel and B. Bozkurt. "Metrical Strength and Contradiction in Turkish Makam Music".
- P. G. Singh. "Auditory Scene Analysis and the Performance and Perception of Tabla Rhythms in Hindustani Music".
Both Turkish and Indian rhythms are very particular. The rhythm in Turkish music revolves around the concept of usul and the one in Indian music is based on the concept of tala. These concepts have been quite well formalized by musicologists but there is practically no computational work on them. Two of the articles presented very preliminary computational work, but there is a lot to be done before the existing musicological formalization can be turned into algorithmic approaches to describe the rhythm of both Turkish and Indian musics. One article presented some work on perceptual aspects of rhythm, which we are not covering within the project.
- F. Font and X. Serra. "Analysis of the Freesound Folksonomy".
- M. Sordo, J. Serrà, G. K. Koduri and X. Serra. "Extracting Semantic Information from on-line Art Music Discussion Forums".
A basic premise in CompMusic is the need to study and characterize the communities, people, that support the music that is being studied. We want to analyze on-line communities and develop methodologies to extract musically relevant knowledge from them. Freesound is an on-line community of people that share and talk about sounds that we have used to start developing methodological approaches of relevance to the project. The first musically relevant on-line community that we have studied has been Rasikas.org, using it to extract musically relevant information for the case of Carnatic music.
Tools for browsing and discovering music:
- M. Sordo, G. K. Koduri, S. Şentürk, S. Gulati and X. Serra. "A Musically Aware System for Browsing and Interacting with Audio Music Collections".
- O. Babacan, C. Frisson and T. Dutoit. "MakamCycle: An Interactive Tool for Browsing Makam Music".
- O. Lartillot and M. Ayari. "An Integrated Framework for Transcription, Modal and Modal and Motivic Analysis of Maqam Improvisation".
- A. Srinivasamurthy and P. Chordia. "A Unified System for Analysis and Representation of Indian Classical Music using Humdrum Syntax".
In CompMusic we are very interested in developing tools and methodologies to be used in practical music applications, especially applications for browsing audio music collections. A relevant functionality is to be able to discover patterns and musical relationships from which a listener can learn to appreciate better the music of a given culture. We presented a paper on the initial application prototype that is developed and other researchers presented systems that are being used in tasks of relevance to CompMusic.
There were two other talks that are outside the proposed organization, one by P. Sarala et al. on "Applause Identification and its Relevance in the Archival of Carnatic Music Audio Recordings" and one by M. Demoucron et al. on "Sculpting the Sound. Timbre-Shapers in Classical Hindustani Chordophones".
A part from the formal talks, the local team put together two very interesting concerts. One was by Bengi Bağlama Üçlüsü, a turkish group dedicated to folk makam music that is lead by Okan Murat Ozturk. The second concert was dedicated to Ottoman music with a group of performers directed by Bekir Reha Sağbaş [picture, video1, video2, video3]. Both concerts included a pedagogical element and the musicians made a special effort to present the pieces and discuss various aspects of their use of makams.
In all academic conferences the informal discussions among the participants during the breaks and social gatherings are as important as the actual formal presentations. This was particularly true in this workshop. Given the focus of the workshop and the shared interests among all the participants, all the breaks and meals were very lively and many interesting discussions emerged. All participants expressed their interest in keeping the communication going after the workshop and in trying to collaborate with other researchers. I will do my best to help with that.
[At the time of writing this post the proceedings are not yet online, but they will be here by the end of Sept. 2012. Also videos of all the talks will be available here by the begining of August 2012.]