In the context of IRISA D5 (Signal, Image & Robotics) department’s scientific seminars, we will be holding a thematic seminar on the modelisation of musical data and signals on Thursday, May 28th 2015.
The modelisation of musical content has long been an area of interest for musicologists, and is an essential step for most of Music Information Retrieval. However, because of the complexity and variety of music, unequivocal models have yet to emerge for high-level concepts such as harmony and structure.
This seminar aims at giving an overview of a few recent approaches coming from the perspectives of engineering science and applied mathematics.
The seminar is open to academic and industrial experts, as well as Ph.D candidates, postdoctorate researchers and colleagues from neighboring research fields who wish to learn about the modelisation of music data, or improve their knowledge of the domain and of its current theoretical and practical challenges.
General program (abstracts further below):
10:15-11:00 – Room AURIGNY
Towards Hierarchical Modelling of Music Structure Using a Minimum Description Grammar
By Corentin Guichaoua (PANAMA Research group / IRISA)
[In teleconference with LABRI / Bordeaux]
14:00-14:45 – Room CRETE
[In French]
Modèles algébriques et topologiques dans le traitement symbolique de l’information musicale
Par Moreno Andreatta (IRCAM – CNRS – UPMC)
14:45-15:30 – Room CRETE
Modeling the perceptive and cognitive dimensions of music structure
By Frédéric Bimbot (PANAMA Research Group / IRISA)
15:30-16:00 – Coffee Break
16:00-17:00 – Room CRETE
Open scientific discussion
—————
Towards Hierarchical Modelling of Music Structure Using a Minimum Description Grammar
Corentin Guichaoua (PANAMA Research group / IRISA)
Modelling music structure, i.e. the organisation of musical elements and their relationships within a piece of music, is an open problem of primary importance in Music Information Retrieval. In this work, we approach music structure description as the inference of a low complexity generative grammar able to account for the music piece, itself represented as a sequence of symbols. Originally introduced for the inference of structure in DNA sequences, Straight-Line Grammars (SLG) form a particular subclass of Context-Free Grammars (CFG) which can be used to model symbolic sequences and to represent them as hierarchical trees. However, SLGs appear to be poorly suited to some particularities of musical patterns, such as segmental regularities, closure substitutions and specific style structures. We propose formal and algorithmic extensions of SLGs as SLEGs (Straight-Line Edition Grammars). Based on a more general minimum description criterion, the SLEG extension allows alterations in the generation step and enables the use of priors in the grammar inference process. We present a diagnostic comparison between the two approaches on the structural segmentation of chord sequences on a set of 20 songs from the RWC-Pop dataset, which indicate that Minimum Description SLEGs offer a promising framework for hierarchical music structure description.
—————
Modèles algébriques et topologiques dans le traitement symbolique de l’information musicale
Moreno Andreatta (IRCAM – CNRS – UPMC)
Depuis une quinzaine d’années la modélisation algébrique des structures et processus musicaux s’est imposée comme un véritable nouveau paradigme en musicologie computationnelle. Etroitement liés à l‘approche transformationnelle en analyse musicale, les modèles algébriques permettent une formalisation à la fois élégante et opérationnelle de nombreux problèmes de nature combinatoire, aussi bien dans le domaine des hauteurs que dans celui des rythmes. Plus récemment, des outils issus de la topologie se sont ajoutés à la palette du musicologue computationnel, en particulier dans la modélisation informatique de l’espace des hauteurs via le Tonnetz et ses généralisations simpliciales. En reprenant quelques résultats de deux thèses de doctorat menées récemment au sein de l’équipe Représentations musicales de l’Ircam (Louis Bigo et Mattia Bergomi), nous allons présenter l’état des recherches sur les structures topologiques appliquées au traitement symbolique de l’information musicale et la classification stylistique automatique en indiquant comment ces outils peuvent être adaptés en vue d’une articulation entre démarche symbolique et approchées basées sur l’analyse du signal audio.
—————
Modeling the perceptive and cognitive dimensions of music structure
Frédéric Bimbot (PANAMA Research group / IRISA)
Given the wide variety of music signals, describing music structure turns out to be a scientific challenge, not only from the algorithmic perspective but also from a conceptual viewpoint.
Approaching the question from an Engineering Science perspective, The “System & Contrast” (S&C) model aims at describing the inner organization of musical segments in terms of : (i) a carrier system, i.e. a sequence of morphological elements forming a multi-dimensional network of self-deducible syntagmatic relationships and (ii) a contrast, i.e. a substitutive element, usually the last one, which partly departs from the logic implied by the rest of the system, and creates a digital modulation with a closural effect.
With a primary focus on pop music, the S&C model provides a framework to describe internal implication patterns in musical segments by encoding similarities and relations between its constitutive elements so as to minimize the complexity of the resulting description. It is applicable at several timescales and to a wide variety of musical dimensions in a polymorphous way, therefore offering an attractive meta-description of different types of musical contents. It has been used as a central component in the creation of a set of annotations for 380 pop songs.
In this talk, we illustrate how the S&C model applies to music structure and we establish its filiation with Narmour’s Implication-Realization model (1990, 1992) and Cognitive Rule-Mapping (Narmour, 2000). We also introduce the Minimum Description Length scheme as a productive paradigm to support the estimation of S&C descriptions.
Originating from an Engineering Science viewpoint, the S&C model establishes promising connections between Music Data Processing and Information Retrieval on the one hand, and modern theories in Music Perception and Cognition on the other hand, together with interesting perspectives in other areas in Musicology.
