Perceptive Strategies in Computational Motivic Analysis: Why and How

Perceptive Strategiesin Computational Motivic Analysis:

Why and How.

Olivier.Lartillot@ircam.fr

www.ircam.fr/equipes/repmus/lartillot

Perceptive Strategiesin Computational Motivic Analysis:

Why and How.The motivic dimension of music, still resisting to a complete and thorough explication, remains one of

the most ambitious domains of interest of music analysis. Music semiology has inspired an ideal of “neutrality”, of the possibility of total independence of the structure to perceptual context. This paradigm has been questioned by competing tendencies that defend the need of a perceptual or even “cognitive” foundation of music analysis. Such dilemma finds a new resonance in today research in automatic musical pattern discovery, which may be considered as a computational inquiry of motivic analysis. Current limitations in this domain seem to stem from an insufficient consideration of the perceptual specificity of musical expression. We propose a general computational model that attempts to mimic music perception. This model relies on two main temporal characteristics of music: chronological direction and short-term selectivity. As a result, musical pattern is defined as an aggregation of successive local intervals. Patterns are induced by analogy between current context and similar past contexts that are reactivated through associative memory. Here, patterns are conceived of as concepts that are actualized in the musical score. This score is represented as a network of notes, which are linked to pattern occurrences that themselves form meta-patterns of patterns. This computational modelling, in process of development as an Open Music library called OMkanthus, aims at offering to musicology a detailed and explicit understanding of music, and suggesting to cognitive science the necessary conditions for musical pattern perception.

Perceptive Strategiesin Computational Motivic Analysis:

Why and How.

Olivier.Lartillot@ircam.fr

www.ircam.fr/equipes/repmus/lartillot

Computational Motivic Analysis

• Automated Music Analysis

• Motivic Analysis– Rudolph Reti– Nicolas Ruwet: Paradigmatic Analysis

• Musical Pattern Discovery– Exact Pattern– Dynamic Programming

Dynamic Programming

ACGGCGTTACGGCAGCGCTGATCGTATCTAGCTAGTCTATGCTAT

ACGGCGTTACGAGCAGCGCTGATCGTATCTAGTAGTCTATGCGAT

CDEFGFEADGAGFEF?

Automated Music Analysis

• Motivic Analysis– Rudolph Reti– Nicolas Ruwet: Paradigmatic Analysis

• Musical Pattern Discovery– Exact Pattern– Dynamic Programming

Perceptual Model?

Music Semiology

ScoreComposer Listener

Poietic Level Neutral Level Esthesic Level

Immanent Structures?

Cultural Knowledge

Cognitive Constraints

Immanent Structures?

Transcendent Structures!

Good patterns

Bad patterns

Automated Music Analysis

• Motivic Analysis– Rudolph Reti– Nicolas Ruwet: Paradigmatic Analysis

• Musical Pattern Discovery– Exact Pattern– Dynamic Programming

Perceptual Model

Perceptive Strategiesin Computational Motivic Analysis:

Why and How.

Olivier.Lartillot@ircam.fr

www.ircam.fr/equipes/repmus/lartillot

Temporal Approach

Temporal Approach

Apprehensive Retention

Apprehensive Retention

Reproductive Remembering

Objectivation

Recognitive Remembering

Recognitive Remembering

Pattern Repetition

Abstract Pattern

Abstract Pattern Tree

Pattern Occurrence Chain

Parallel Patterns

Architecture• loop for note in score

– memorize new retentions– develop current expected occurrences– develop current unexpected occurrences– develop current objectivations– find new objectivations

OMkanthus 0.1