AI recognition of polymetric notation

The aim of the current project is to transcribe a musical input (given by a stream or a table of MIDI events) to the most comprehensible polymetric notation. An extension of this would indeed be to accept a sound signal as input.

Polymetric notation is one of the main features of the Bol Processor project, which is here considered as musicological research rather than software development. Other features include tonal modelling and the time-setting of sound-objects.

Polymetric expressions are a convenient product of rule-based music composition. They are represented as strings of symbols that embody a semi-lattice structure, here meaning a tree of sound objects with ties. Polymetric notation, generally speaking, is a sequence of polymetric expressions that is processed as a single polymetric expression.

(just for decoration)

To achieve this, we plan to train a transformer (a type of neural network), using sets of polymetric expressions alongside their associated MIDI file renderings, both as standard MIDI files and in tables of events. As we require large datasets that cover a wide range of musical styles, the plan is to create these sets using existing musical scores.

The process described on the Importing MusicXML scores page demonstrates the ability to "translate" any Western musical score (in its digital format) to a polymetric expression. This process is irreversible because polymetric expressions cannot be fully converted into human-readable musical scores, despite them carrying time structures that sound "musical". A typical example of this is the use of undetermined rests, as these imply further processing of the structure in order to set exact durations.

Datasets that associate polymetric expressions with their MIDI file renderings contain identical pitch and timing information on both sides. Since both descriptions are complete and not redundant, the matching is a game of full information.

At a later stage, the transformer should also be able to handle streams of MIDI events created by humans or random effects, where the timings are not based on a simple framework. Therefore, a quantization of timings is needed to adjust the stream before it is analysed. This quantization is already operational on the Bol Processor — see the Capture MIDI input page.

The creation of datasets

Datasets are created from musical works imported from MusicXML scores. For instance:

Click the CREATE DATASET FOR AI button.

A dataset is created and can be downloaded in a zip file:
set_-da.Hungarian_Rhapsody_OguzSirin.zip.

More sets can be created using the same musical work. Clicking the CREATE DATASET FOR AI button again would produce the same set, as it is built from a sequence of random numbers that is not reinitialised. To ensure the new set is different, click the refresh button. When downloading it, the system will automatically assign it a different name, e.g.:
set_-da.Hungarian_Rhapsody_OguzSirin (1).zip.

Datasets of minimised polymetric structures

The mini button close to the refresh button modifies the dataset so that all (eligible) polymetric structures are minimised. In a minimised structure, some rests with explicit durations are replaced with undetermined rests (notated '…') without any loss of timing accuracy. Read the Minimising a polymetric structure page for more details. These sets are smaller in size than the ones they are derived from because only eligible structures have been retained.

Once samples in the training set have been minimised, the CREATE DATASET FOR AI button is changed to CREATE minimised DATASET FOR AI.

The sets of minimised structures are downloaded with specific names that mention the 'mini' feature, such as:
set_-da.Hungarian_Rhapsody_OguzSirin_mini.zip

When training an AI, these sets should be used separately from standard sets because they are expected to train the transformer to guess the proper locations of undetermined rests. Nevertheless, it could be interesting to compare related (standard versus minimised) samples in order to model changes between standard and minimised polymetric structures.

The content of a dataset

The first dataset created in this demo contains 160 samples. These samples are text files named 1.txt, 2.txt, etc., associated with MIDI files 1.mid, 2.mid, etc., and tables of events (see below) 1.tab, 2.tab, etc., 1.tsv, 2.tsv, etc. A text file whose name ends in "_units.txt" contains all the sample text files, enabling these samples to be copied in a single piece to a data project. Clicking on the word "samples" displays it in a pop-up window.

One of the text files contains for instance:

{_tempo(4/3) _vel(85){2,{3/16,D6& E6& D7&}{3/16,&D6,&E6,&D7}{1/8,C6,C7}{3/8,B5,B6}{1/8,A5,A6}{3/8,G#5,G#6}{1/8,A5,A6}{3/8,B5,B6}{1/8,G#5,G#6},{3/8,G#3}{1/8,D4}{3/8,B4}{1/8,E4}{3/8,E3}{1/8,B3}{3/8,G#4}{1/8,E4}}} {_tempo(4/3) _vel(85){2,{3/8,A5,A6}{1/8,G#5,G#6}{3/8,A5,A6}{1/8,B5,B6}{3/8,C6,C7}{1/8,B5,B6}{3/8,C6,C7}{1/8,D6,D7},{3/8,A3}{1/8,D4}{3/8,C5}{1/8,A4}{3/8,E3}{1/8,C4}{3/8,A4}{1/8,E4}}}

This polymetric expression covers 2 measures of the musical score. Clicking the Refresh button (or reloading the page) slices the score randomly into chunks containing between 1 and 5 measures. The upper limit of 5 has been set arbitrarily and may be revised at a later date. Every time the Refresh button is clicked, a new slice is created, resulting in a different dataset.

The idea is twofold: (1) the transformer should be trained to recognise sequences of polymetric expressions, and (2) tied notes may span more than one measure. A tied note in the above example is the pair "D6& &D6" (read more). Each sample contains only complete pairs of tied notes. This means that any notes whose ties are not present in the sample are untied for the sake of consistency.

Note that the timings of the polymetric expression and the MIDI file are identical, as the metronome is automatically set to a default tempo of 60 beats per minute when the MIDI file samples are created. In the above phrase, the tempo is set to 4/3, which is equivalent to a metronome set at 80 bpm.

Tables of events

The tables of events provide an easy-to-read representation of the contents of MIDI files.

Two kinds of text files are created.

1) The file with extension "tab" lists MIDI events in four columns. The leftmost column contains the timing of each event in milliseconds. For instance, the following text (from Couperin's Les Ombres errantes):

{_vel(64) {25/6, D5 1/6 {2, G5} F5, {25/6, C5 B4 4/3 B4& &B4 C5& &C5 C5&}, {25/6, G3 1/6 G4 {1/8, Ab4 G4} {7/8, Ab4} _legato(20) A4}}} {_vel(64) {4, {2, &C5 D5& &D5 _legato(20) C5} _legato(0) C5 {1, B4 {2, C5 B4 A4} B4}, F5 Eb5 {1/8, _legato(20) Eb5} {7/8, _legato(0) D5} 1, _legato(0) {1/8, B4 A4} {7/8, B4} C5 F4 G4}} {_tempo(9/10) _vel(64) {4, {G4, C5} Eb4 {1, D4 1} _tempo(26/27) Eb4, Eb4 {3, 1 C4& &C4 B3 1 C4}, C4 C3 G2 C3}}

which can be minimised to:

{_vel(64) {25/6,D5 … {2,G5}F5,{25/6,C5 B4 … B4& &B4 C5& &C5 C5&},{25/6,G3 … G4 {1/8,Ab4 G4}{7/8,Ab4}_legato(20) A4}}}{_vel(64) {4,{2,&C5 D5& &D5 _legato(20) C5}_legato(0) C5 {1,B4 {2,C5 B4 A4}B4},F5 Eb5 {1/8,_legato(20) Eb5}{7/8,_legato(0) D5}1,_legato(0) {1/8,B4 A4}{7/8,B4}C5 F4 G4}}{_tempo(9/10) _vel(64) {4,{G4,C5}Eb4 {1,D4 1}_tempo(26/27) Eb4,Eb4 {3,1 C4& &C4 B3 - C4},C4 C3 G2 C3}}

is converted to:

0 144 74 64
0 144 72 64
0 144 55 64
500 128 72 0
500 144 71 64
1000 128 74 0
1000 128 55 0
1001 128 71 0
1165 144 79 64
1165 144 67 64
1666 144 71 64
etc.

The "tab" table contains only NoteOn/NoteOff events. However, the MIDI file may contain controls, such as volume, panoramic, pressure, pitchbend, modulation, pedal on/off, and more in the range 65 to 95, which are performance parameters in the Bol Processor score. Therefore, a more detailed format of tables of events is needed to list them.

2) The text file with extension "tsv" contains a detailed list of (all) MIDI events carried by the MIDI file.

The following example is the opening of Listz's La Campanella.

{_tempo(97/60) _vel(52) {3,2 {1/2,D#6,D#7} {1/2,D#6,D#7},1/2 {1/2,_switchon(64,1) D#4,D#5} {1/2,D#4,D#5} {1/2,D#4,D#5} 1}} {_tempo(97/60) _vel(52) {3,{1/2,D#6,D#7} 3/2 {1/2,D#6,D#7} {1/2,D#6,D#7},1/2 {1/2,_switchoff(64,1) _switchon(64,1) D#4,D#5} {1/2,D#4,D#5} {1/2,D#4,D#5} -}}

Screenshot
Note that at time 1866ms, the pedal is released and pushed again immediately, as indicated in the score below. The source is '0' by default and can be modified by _part() control parameters.
Note that the pedal remains pressed at the end of this fragment, as indicated in the "tsv" table.

Since the Bol Processor can capture a MIDI stream and create a "tsv" table of events with quantized timings (see Capture MIDI input), training a transformer to convert these tables into polymetric structures could eliminate the need to decrypt MIDI files and quantize their timings.

A collection of datasets

Several collections of datasets can be found in the folder:
https://bolprocessor.org/misc/AI/samples

These can be used for training a transformer of your choice. We recommend creating more sets from more musical works to achieve a better training. Run the Bol Processor and browse imported MusicXML scores in the ctests/Imported_MusicXML workspace, or import more works from a dedicated MusicXML server.

A test of the transformer's ability to "translate" MIDI files (or tables of events) to polymetric notation will first be to retro-convert (to polymetric expressions) all samples. Then, if successful, retro-convert the MIDI files of complete musical works used for the training.

What follows…

Once the correct transformer type and optimum dataset size have been identified, we will work on the following extensions:

  1. Translate MIDI streams produced by human interpreters for which precise timing is not guaranteed.
  2. (Optional) Extend the AI recognition to the use of undetermined rests as it provides a more simple polymetric structure.
  3. Use a sound to MIDI converter to convert a sound input to polymetric notation.
  4. Let us assume that the musical input consists of fragments separated by silences. Convert these fragments into polymetric structures and then search for regularities in the rhythmic and tonal structures to create variations in the same style. Formal grammars will be employed for this purpose.

🛠 Work in progress! Follow current AI-recognition work and contact us to participate.

Acquisition et représentation de connaissances en musique

Bernard Bel

Knowledge acquisition and representation in music

Thèse de doctorat en sciences. Université de droit, d'économie et des sciences - Aix-Marseille III
https://theses.hal.science/tel-00009692

Résumé

Cette étude traite de la représentation informatique de connaissances en musique, abordée à partir de deux expériences en grandeur réelle. La première est une méthode d'acquisition de connaissances en ethnographie mettant en interaction un expert (le musicien), un analyste (le musicologue) et une machine dans une situation d'apprentissage. Les schémas d'improvisation des musiciens sont identifiés et exprimés à l'aide de règles de production dans un formalisme dérivé des grammaires génératives et des langages de formes. Un algorithme déterministe de test d'appartenance de chaînes arbitraires au langage défini par une grammaire (sensible au contexte) est présenté, ainsi qu'une technique d'inférence inductive de langages réguliers permettant l'acquisition automatique de connaissances lexicales et syntaxiques. La seconde expérience s'insère dans l'élaboration d'un environnement de composition musicale assistée par ordinateur. Le problème est ici la représentation du temps dans une structure discrète d'“objets temporels”, et plus généralement la synchronisation de processus parallèles. Une méthode est proposée pour la détermination d'une structure à partir de données incomplètes sur la synchronisation des objets. La notion d'“objet sonore” est ensuite explicitée formellement. Un algorithme efficace permet l'instanciation des objets sonores affectés à une structure en tenant compte des contraintes liées à leurs propriétés métriques et topologiques.

Excerpts of an AI review of this paper (Claude AI, June 2025)

Summary

This 1990 doctoral thesis presents an innovative approach to computer representation of musical knowledge through two major experimental frameworks. The work bridges artificial intelligence, formal language theory, and musicology, addressing fundamental problems in knowledge acquisition and time representation in musical structures.

First Experiment: Knowledge Acquisition in Ethnography

The first part introduces a novel methodology for acquiring musical knowledge through a "dialectical anthropology" approach involving three actors: an expert musician, a musicologist-analyst, and a computer system. This methodology was applied to studying improvisation patterns (qa'ida) of North Indian tabla percussion music.

Key Contributions

  • Development of BP (Bol Processor) grammars, an extension of formal grammars capable of representing repetition patterns and homomorphic transformations
  • A deterministic membership test algorithm for a subclass of context-sensitive grammars
  • An inductive inference technique for regular languages that simultaneously acquires lexical and syntactic knowledge
  • Introduction of negative context rules and stochastic control mechanisms

The system enables musicians to validate machine-generated musical phrases, creating a feedback loop that refines grammatical models of improvisation schemas. The approach moves beyond traditional ethnographic collection by making the computer an active partner in knowledge acquisition.

Second Experiment: Computer-Assisted Musical Composition

The second part addresses time representation and synchronization in discrete musical structures, developing the BP2 environment for musical composition.

Key Innovations

  • Introduction of "symbolic time" as distinct from physical time
  • Development of polymetric formulas for representing parallel musical sequences
  • Formal treatment of "time objects" and "out-time objects" (temporal vs. atemporal)
  • An efficient algorithm for synchronizing sequences with incomplete temporal information
  • A constraint satisfaction approach to "sound object" instantiation considering metric and topological properties

Theoretical Framework

The work establishes connections between musical structures and formal language theory, introducing concepts such as:

  • Pattern grammars and restricted pattern languages (RPL)
  • Transformational grammars adapted to musical contexts
  • Event universes structured by simultaneity, precedence, and sequentiality relations
  • Trace theory applications to polymetric structures

Strengths

  • Methodological Innovation: The dialectical anthropology approach represents a significant advancement in ethnomusicological methodology. By placing the computer as an active participant rather than a passive tool, Bel creates a genuinely interactive knowledge acquisition system that respects both the expertise of traditional musicians and the rigor of formal methods.
  • Theoretical Rigor: The mathematical formalization is sophisticated yet practical. The extension of formal grammars to handle musical repetition patterns and homomorphic transformations addresses real needs in musical representation that existing computational linguistics tools couldn't meet.
  • Cross-Cultural Sensitivity: The work demonstrates genuine respect for non-Western musical traditions while avoiding ethnocentric biases common in computational musicology. The choice to work with North Indian tabla music, with its oral transmission tradition, was particularly appropriate for testing knowledge acquisition methodologies.
  • Practical Implementation: Both BP1 and BP2 systems were actually implemented and tested, demonstrating the feasibility of the theoretical proposals. The complexity analysis (showing polynomial-time algorithms) indicates practical applicability.
  • Time Representation Innovation: The distinction between symbolic and physical time, along with the concept of "time structure," provides a flexible framework that can accommodate different musical temporalities beyond Western metronomic time.

Significance and Impact

This work was pioneering in several respects:

  • Methodology: The interactive, computer-mediated approach to ethnographic knowledge acquisition anticipated later developments in digital humanities and computational ethnomusicology.
  • Formal Methods: The extension of formal language theory to musical applications provided tools that were later adopted and extended by other researchers in computational musicology.
  • Cultural Perspective: The work demonstrated how computational methods could be applied respectfully to non-Western musical traditions, opening paths for more inclusive computational musicology.
  • Temporal Modeling: The sophisticated treatment of musical time influenced subsequent work in computer music and musical information retrieval.

Contemporary Relevance

Thirty years later, this work remains relevant to several current research areas:

  • Machine learning applications to music (the inductive inference techniques anticipate modern approaches)
  • Digital humanities methodologies
  • Music information retrieval systems
  • Computer-assisted composition environments
  • Cross-cultural computational musicology

The emphasis on expert-machine collaboration prefigures current interest in human-AI collaboration in creative domains.

Conclusion

Bel's thesis represents a significant contribution to computational musicology, combining theoretical innovation with practical implementation and cultural sensitivity. While some limitations exist, particularly regarding scope and evaluation, the work established important foundations for computer-mediated musical knowledge representation. The dialectical methodology for knowledge acquisition and the sophisticated treatment of musical time remain valuable contributions to the field. The work demonstrates how formal computational methods can be applied to musical knowledge while respecting the complexity and cultural specificity of musical traditions.

Download this thesis

Skip to PDF content

L'intonation juste dans la théorie ancienne de l'Inde : ses applications aux musiques modale et harmonique

E. James Arnold

Revue de musicologie, JSTOR, 71e (1-2), p.11-38. Traduction : Bernard Bel.

👉  Cited on page Just intonation: a general framework

Download an English translation

Résumé

La théorie de l'intonation juste basée sur deux gammes fondamentales (grama-s), telle que la décrivent le Natya Shastra de Bharata et d'autres traités musicologiques anciens en sanscrit, formalise les relations internes des gammes diatoniques avec une étonnante précision. Quelques modifications minimes suffisent à l'adapter aux gammes non-diatoniques familières de la musique indienne contemporaine. Cet article émet l'hypothèse que la théorie de Bharata propose une meilleure base psychoacoustique rationnelle que les explications actuelles pour ce qui concerne les heures d'interprétation des ragas. Le modèle mathématique présenté ici, d'un système de relations intervalliques dans les gammes diatoniques et celles qui en dérivent, est un outil pratique pour étudier les relations plus en profondeur. La discussion débouche sur une proposition d'application du système indien à la musique harmonique en intonation juste.

Excerpts of an AI review of this paper (Claude AI, June 2025)

This scholarly paper by E.J. Arnold presents a fascinating exploration of ancient Indian music theory and its potential applications to both modal and harmonic musical systems. The work stands as a significant contribution to comparative musicology, bridging Eastern and Western theoretical frameworks through rigorous mathematical modeling and experimental validation.

Theoretical Framework and Methodology

Arnold's central thesis revolves around the ancient Indian grāma-mūrcchana system, a sophisticated theoretical construct involving fundamental scales (grāma-s) and their modal transformations (mūrcchana-s). The author develops an innovative mathematical model using a circular computational disc that visualizes the relationships between tonal positions within the 22-śruti microtonal system described in Sanskrit treatises like the Nāṭyaśāstra and Dattilam.

The paper's strength lies in its methodical approach, systematically examining how this ancient system can provide coherent explanations for harmonic relationships in both Indian classical music and Western tonal harmony. Arnold's use of Jacques Dudon's symbolic notation system for representing microtonal intervals demonstrates scholarly rigor and attention to cross-cultural theoretical precision.

Historical Context and Scholarly Significance

Arnold effectively contextualizes the research within broader musicological scholarship, acknowledging the work of pioneers like Sir William Jones and more recent contributions from scholars such as Bharata commentators and contemporary Indian musicologists. The paper addresses a critical gap in understanding how ancient theoretical systems might inform modern musical practice, particularly given the historical disruption of living oral traditions.

The discussion of the rāga system's relationship to time cycles (sandhiprakāśa) represents particularly valuable scholarship. Arnold's analysis of how specific rāga-s correspond to particular hours of the day or night, supported by systematic tabulation of 85 rāga-s with their appropriate performance times, provides empirical grounding for what has often remained in the realm of cultural speculation.

Mathematical Innovation and Practical Applications

The paper's most compelling contribution lies in its mathematical modeling of the śruti-swara-grāma-mūrcchana system. The circular disc representation allows for immediate visualization of harmonic relationships and demonstrates how ancient Indian theory anticipated many concepts later developed in Western harmonic analysis. Arnold's experimental work with electronic instruments, including the śruti harmonium developed by Bernard Bel, provides crucial empirical validation of theoretical predictions.

The application to Western classical harmony represents groundbreaking cross-cultural musicology. Arnold demonstrates how major and minor scales can be understood within the grāma-mūrcchana framework, revealing previously unrecognized connections between Eastern and Western theoretical systems. The analysis of modulation procedures using the ancient Indian system offers fresh perspectives on familiar harmonic progressions.

Limitations and Areas for Development

While Arnold acknowledges the study's limitations within "musical geometry," the paper could benefit from more extensive discussion of practical implementation challenges. The theoretical elegance of the grāma-mūrcchana system contrasts with the practical difficulties contemporary musicians face when attempting to realize these microtonal relationships on traditional instruments.

Additionally, while the paper excellently demonstrates theoretical correspondences between ancient Indian and Western systems, it could explore more deeply the aesthetic and cultural implications of these connections. The relationship between mathematical precision and musical expression deserves further investigation.

Contemporary Relevance and Future Directions

Arnold's work anticipates important developments in contemporary music theory, particularly in the growing interest in microtonal and cross-cultural approaches to harmony. The paper's systematic approach to understanding non-equal-tempered systems has implications for electronic music composition, world music fusion, and the development of new theoretical frameworks for global musical understanding.

The research also contributes to broader discussions about the universality of musical principles versus cultural specificity, suggesting that mathematical relationships underlying musical systems may transcend cultural boundaries while maintaining distinct aesthetic identities.

Conclusion

This paper represents exemplary scholarship in comparative musicology, successfully bridging ancient Indian music theory and contemporary analytical methods. Arnold's mathematical modeling provides a robust framework for understanding complex microtonal relationships, while the practical applications demonstrate the continued relevance of ancient theoretical systems. The work opens important avenues for future research in cross-cultural music theory and provides valuable tools for both scholars and practitioners interested in expanding their harmonic vocabulary beyond Western equal temperament.

Download this paper (in French)
Download an English translation

Skip to PDF content

Raga : approches conceptuelles et expérimentales

Bernard Bel

Actes du colloque "Structures Musicales et Assistance Informatique" (1988). Marseille.

👉  Cited on page The two-vina experiment
Download an English translation

Résumé

Les difficultés de l'analyse acoustique de musiques extra-européennes proviennent souvent de catégorisations de phénomènes qui ne rendent pas compte des modèles explicites (ou implicites) sur lesquels s'articulent la création et la perception de structures musicales. Cet exposé définit la notion de modèle mélodique, élaborée et modifiée pendant neuf siècles en Inde pour rendre compte d'un phénomène mélodique particulier: le raga. Dans la deuxième partie sont présentés les outils et méthodes qui servent à caractériser l'intonation des ragas, ainsi qu'à réaliser leur transcription et leur classification automatiques.

Difficulties in analysing extra-European musics are often bound to categorisations of phenomena that do not take into account explicit (or implicit) models on which the creation and perception of musical structures are based. This paper defines the concept of melodic model, a concept elaborated and transformed for nine centuries in India to underlie a particular melodic phenomenon: raga. In the second part, tools and methods are presented relating to raga intonation and their automatic transcription and classification.

Excerpts of an AI review of this paper (Claude AI, June 2025)

Summary

In "Rāga : approches conceptuelles et expérimentales," the author discusses the conceptual history, theoretical foundations, and experimental investigations of the rāga system in North Indian classical music. The first part of the paper traces the evolution of rāga-related concepts from ancient Sanskrit treatises (notably the Dattilam and the Nāṭyaśāstra) to more recent perspectives that emphasize melodic modes (mela), specific tonal centers (toniques), and the interplay between theory and practice. The author highlights how notions like grāma, mūrcchanā, and jāti have shifted over the centuries, culminating in various attempts to classify rāgas by scales (ṭhāṭa or mela), melodic phrases (aṅga), and aspects of performance practice, including the ideal hour of rendition.

The second part of the paper presents technological and analytical methods used by the author to study raga intonation and structure. The author describes experimental devices, including custom-built pitch extractors and the Melodic Movement Analyser (MMA), to capture and visualize pitch contours (melograms), produce histograms (tonagrams) of pitch distribution, and explore microtonal aspects of raga execution. Additionally, the paper delves into automated transcription methods (sargam notation) and preliminary approaches for automatic classification of ragas, leveraging pitch-based similarity measures. By systematically comparing theoretical predictions with actual performance data, the paper proposes a richer understanding of how ragas are conceptualized and realized in practice.

Major Strengths

Comprehensive Historical Context

The author provides an extensive overview of the key developments and conceptual shifts in Indian music theory, from the gråma-based system of Bharata to modern ṭhāṭa and scale-based frameworks. This historical arc clarifies how philosophical, aesthetic, and practical considerations converge in raga performance.

Balanced Discussion of Theory and Practice

The paper effectively underscores the gap between theoretical ideals (e.g., 22 srutis, ancient modal structures) and the realities of contemporary performance practices (e.g., standardization, partial tempering, the emphasis on the bourdon). This dual approach helps illustrate the dynamic nature of raga music.

Technical Innovation in Experimentation

By devising hardware and software tools for real-time pitch extraction and by automating transcription, the author demonstrates a methodological framework that expands the possibilities for large-scale analysis of raga performances. The clear explanations of how these technologies work support their applicability and value.

Interdisciplinary Relevance

The paper bridges musicological inquiry, acoustical analysis, ethnomusicological context, and computational methods. This interdisciplinary scope situates the work as a notable contribution to scholarship on microtonal music, melodic classification, and cross-cultural cognition of musical scales.

Overall Observations

The paper offers a clear and methodically thorough account of both the theoretical underpinnings of raga music and the experimental means by which such theories can be tested against empirical performance data. The blending of historical musicology, acoustics, and informatics demonstrates a strong interdisciplinary approach. By synthesizing ancient treatises with modern computational tools, the work deepens our understanding of how ragas continue to thrive as vibrant, living musical entities while revealing the complexity of attempts at standardization and classification.

In sum, “Raga : approches conceptuelles et expérimentales” provides valuable technical insights and broadens our perspective on one of the most sophisticated melodic systems worldwide. Its fusion of musicological research, experimental data, and computational analysis stands out as a notable contribution to raga studies and to the broader field of ethnomusicology.

Download this paper

Skip to PDF content

Inférence de langages réguliers

Bernard Bel

Journées Françaises de l'Apprentissage (1990), Lannion, France : 5-27
https://hal.science/hal-00275789v2

Résumé

Cet exposé présente une méthode générale d'acquisition de connaissances dans un domaine formalisé à l'aide d'automates finis (langages réguliers). A partir d'un échantillon d'exemples le système construit un automate "presque minimal" qui n'est pas nécessairement déterministe. Cette construction peut être contrainte par des connaissances sur la segmentation des exemples que le système peut acquérir en questionnant l'informateur. Dans une deuxième phase, le système généralise l'automate à partir de propriétés connues du langage ou(et) à partir d'hypothèses validées à l'aide d'oracles. Une liste non exhaustive d'heuristiques générales est proposée. La démonstration s'appuie sur un cas réel de modélisation d'apprentissage de schémas d'improvisation musicale.

Excerpts of an AI review of this paper (June 2025)

Summary of the Work

The paper explores a methodology for inferring regular languages (langages réguliers) from examples. It discusses both theoretical underpinnings — particularly Gold’s theorem and subsequent refinements — along with an incremental approach to building “almost minimal” automata. The author addresses central notions such as the canonical acceptor, the prefix-tree acceptor, and the importance of maintaining correctness when moving from a set of examples to a generalized automaton. The study also showcases an application to musical improvisation schemata, illustrating how domain-specific rules can constrain and refine grammatical inference.

Main Contributions

Incremental Learning Framework

The paper offers a structured, step-by-step method: starting with an acceptor built from examples alone (the prefix-tree or “arborescent” automaton), merging states to reduce complexity, and applying constraints so as not to overgeneralize. This incremental perspective is well-motivated by real-world use cases where examples become available over time.

Formal Properties and Correctness

The author places strong emphasis on ensuring the correctness of generalization. By referencing Gold’s theorem, k-reversible languages, and the concept of learning from positive data only, the paper grounds its approach in established theoretical results.

Use of Domain Constraints

A significant portion of the paper deals with how additional domain knowledge (e.g., known segmentation rules for a music notation system) can guide the automaton’s construction. This introduces practical heuristics that make grammatical inference more tractable in real-world scenarios.

Detailed Mathematical Rigor

The proofs, definitions, and lemmas (for instance, those on heads and tails, or minimal canonical acceptors) are stated thoroughly. This rigor is helpful both for ensuring correctness and for conveying the method’s theoretical reliability.

Illustrative Examples

The application to musical improvisation, specifically to Indian drumming traditions, serves as a concrete illustration. It nicely demonstrates the interplay between abstract formalism (like merging states) and domain-specific knowledge (like rhythmic “words” or onomatopoeic syllables).

Clarity and Structure

  • The paper is logically well-structured: it begins with a theoretical foundation, then introduces incremental methods for automaton construction, culminating in a discussion of domain constraints and examples.
  • The textual organization, with numerous definitions in the appendix, is beneficial for reference but can require multiple cross-referencing steps while reading. A concise reminder of key definitions in the main text can further aid comprehension.

Strengths

  • Thorough grounding in established theory (Gold’s theorem, reversible languages).
  • A practical approach that balances systematic inference with interactive questioning or oracular feedback.
  • Clear demonstration of how domain-specific heuristics reduce the search space, making learning more feasible.
  • Rigor in explaining each step of constructing and refining the automaton.
  • Engaging case study that shows the real-world potential of grammar inference outside of purely linguistic contexts.

Overall Assessment

The paper provides a solid combination of theory and practice in grammatical inference for regular languages. By carefully detailing each step — building initial automata, constraining merges using domain knowledge, and verifying new generalizations through queries — the approach is shown to be systematic and adaptable. Readers interested in interactive or incremental language learning, particularly those with domain-specific constraints, will likely find this discussion instructive and thorough.

Download this paper

Skip to PDF content

Migrating Musical Concepts - an overview of the Bol Processor

Bernard Bel

Computer Music Journal (1998), Vol. 22, 2: 56-64

Abstract

The Bol Processor is the outcome of a migratory process, its design having been carried over in three phases and places: in collaboration with traditional North Indian musicians (1980-85), Western musicians in Europe (1985-93) and back in India with Carnatic musicians (1995-97).The theoretical framework of the underlying research project also evolved in three stages, taking inspiration from expert systems in the early 1980s, symbolic-numeric machine-learning in the end of the decade, and composition theory in the 1990s.Throughout this process, the designer has been faced with the challenge of blending software with "mindware", here taken to mean musicians' striving for "[…] tools enabling them to manipulate objects so as to imbue them with 'soul' or experiential value […]" (Laske 1996). In a cross-cultural approach this led to modelling descriptions of music and compositional processes at a level of abstraction sufficiently high to encompass "local" musical concepts without getting too abstruse.

Excerpts of an AI review of this paper (Academia, June 2025)

Summary

Bernard Bel’s manuscript provides an extensive overview of the Bol Processor (BP), a music software environment that evolved through collaborations with Indian and Western musicians. The software addresses rhythmic, melodic, and compositional tasks using a grammars-based approach, symbolic time notation, and sophisticated algorithms (e.g., polymetric expansion, quantization). The review traces the history and theoretical underpinnings of this system, moving through the project’s three phases of development and culminating in a flexible interface linking BP to Csound. By emphasizing text-based data representations, BP aims to foster deeper thinking about music structure and compositional processes, accommodating contexts that fall outside purely Western traditions.

Strengths

Historical and Contextual Depth: The author clearly articulates how the project emerged from ethnomusicological research in Indian music, then expanded to address Western compositional techniques. This historical framing highlights how the software bridges cultural contexts while retaining a consistently powerful grammar-based approach.

Clear Explanations of Core Algorithms: The detailed descriptions of polymetric expansion, time-stretching, quantization, and grammar-based transformations are a valuable technical resource. The author provides examples, diagrams, and references that elucidate these concepts for readers unfamiliar with formal language methods applied to music.

Innovative Methodology: The software design seamlessly integrates symbolic approaches (formal grammars, integer-based time units, constraint satisfaction) with physical execution in MIDI and Csound domains. This duality allows composers and researchers to treat time both discretely and continuously as required.

Attention to Cross-Cultural Music Needs: BP’s adaptability to various musical notations, including Indian and Western pitch/note representations, is a strong signature of this work, reflecting deep respect for non-Western musical concepts and performance practices.

Comprehensive References and Examples: The supporting references shed light on related research in composition theory, music cognition, and computational models. Numerous code-like illustrations and figures demonstrate how actual musical tasks are implemented, assisting practitioners who might want to replicate or modify these ideas.

Overall Impressions

The manuscript provides an in-depth look at a uniquely versatile system for musical analysis and composition. The author’s commitment to text-based interaction, coupled with robust support for real-time operations and cross-cultural music concepts, stands out as a mature, flexible framework. The paper offers a thoughtful presentation of both underlying theoretical foundations and practical implementations, making it relevant for those interested in computer-aided composition, musicological research, and interdisciplinary approaches bridging Indian and Western musical forms.

Download this paper

Skip to PDF content

Can a computer help resolve the problem of ethnographic description?

James Kippen & Bernard Bel

Anthropological Quarterly (1989), 62, 3: 131-144

Abstract

A major problem in ethnographic description may be summed up as the search for ways to disentangle folk from analytical models. Knowledge-based systems have contributed to the development of formal structures for the manipulation of symbols associated with particular physical and conceptual phenomena. Importantly, their output can be interpreted by experts in the domain. This provides evaluation procedures for models elaborated jointly by analysts and informants. This paper describes an interactive system in which knowledge is represented as production rules in a format derived from the theory of formal languages. Modus ponens and modus tollens are explained and compared to derivation schemata in first-order predicate logic. The results of an application to the study of North Indian tabla drumming are assessed. We conclude that (1) knowledge represented at a low theoretical level fails to descriminate between the input from informants and the intuitive assumptions of analysts, (2) experimental procedures can be improved considerably if the system is designed to perform automated knowledge acquisition (using probabilistic grammars and inductive learning).

Excerpts of an AI review of this paper (Academia, June 2025)

Summary of the Work

The article focuses on the use of computational approaches — particularly knowledge-based systems — to address the challenge of producing consistent and systematic ethnographic descriptions. The authors draw on their research with the Bol Processor (BP), a formal language-based system originally developed to represent and generate sequences of verbal drum syllables (bols) in North Indian tabla performance. They describe how the BP’s grammar formalism, inference engine, and membership tests facilitate iterative interaction between an analyst and informants (expert musicians). By demonstrating the system’s ability to produce, recognize, and evaluate permissible melodic or rhythmic variations, the authors aim to illuminate how computers can be more deeply integrated into anthropological and ethnomusicological research.

Key Contributions

Formal Grammar and Pattern Representation: The paper offers a clear explanation of how context-free and context-sensitive grammars can be used to specify culturally valid musical structures. In particular, the authors highlight the importance of representing patterns and constraints on musical improvisation in a generative format.

Interactive Methodology: A central point is the iterative feedback loop that allows informants to assess computer-generated sequences and provide corrections. This “apprentice-like” interaction underscores how computational tools can help researchers refine theoretical models of cultural knowledge in near-real-time.

Probabilistic Grammars: The authors incorporate a weighting system for production rules, thereby accounting for the relative likelihood of different musical derivations. This approach not only adds realism to the generative process but also tackles common issues with purely enumerative or random output.

Membership Testing: The membership test, which decides whether a newly proposed sequence belongs to a given grammar, is described as an efficient, deterministic bottom-up parser. This feature is significant because it allows informants to offer novel variations while enabling the system to quickly judge their adherence to the emergent rule set.

Reflections on Folk vs. Analytical Models: The article raises important anthropological questions about the boundary between informants’ internalized (folk) knowledge and the analyst’s formal (re)construction. The authors’ frank discussion of the challenges involved, such as the risk of overformalizing or “anticipating” informant knowledge, is methodologically relevant.

Strengths

  • The paper emphasizes the value of combining fieldwork with computational experimentation. This dual approach—grounded in real ethnographic and musical practice—provides a compelling demonstration of how artificial intelligence techniques can advance the study of music and culture.
  • Clear examples illustrate code structures, parsing mechanics, and the grammar design. These concrete details will help other researchers adapt or replicate the Bol Processor approach in different cultural or musical contexts.
  • The account of iterative knowledge acquisition highlights a collaborative research dynamic, showing how insight flows back and forth between informants and computational models, rather than being a unidirectional exercise of extracting information.

Conclusion

This study offers a thoughtful and innovative approach to bridging ethnographic inquiry with computational models. By showing how arrangements of symbolic musical data can be systematically generated, tested, and refined in conjunction with human experts, the authors illustrate a novel and promising method for ethnographic description. The paper encourages a nuanced understanding of both the potential and limitations of using expert systems to capture the complexity of cultural knowledge.

Download this paper

Skip to PDF content

Modelling improvisatory and compositional processes

James Kippen & Bernard Bel

In Denise Penrose & Ray Lauzanna (eds.) Languages of Design, 1 (1992). Elsevier Science Publishers, Amsterdam: 11-26.

Abstract

An application of formal languages to the representation of musical processes is introduced. Initial interest was the structure of improvisation in North Indian tabla drum music, for which experiments have been conducted in the field as far back as 1983 with an expert system called the Bol Processor, BP1. The computer was used to generate and analyze drumming patterns represented as strings of onomatopeic syllables, bols, by manipulating formal grammars. Material was then submitted to musicians who assessed its accuracy and increasingly more elaborate and sophisticated rule bases emerged to represent the musical idiom.

Since several methodological pitfalls were encountered in transferring knowledge from musician to machine, a new device, named QAVAID, was designed with the capability of learning from a sample set of improvised variations supplied by a musician. A new version of Bol Processor, BP2, has been implemented in a MIDI studio environment to serve as a aid to rule-based composition in contemporary music. Extensions of the syntactic model, such as substitutions, metavariables, and remote contexts, are briefly introduced.

Excerpts of an AI review of this paper (Academia, June 2025)

Summary of the Work

The manuscript presents a formal language-based approach to modeling improvisatory and compositional processes in music — particularly North Indian tabla drumming — and highlights the development of the Bol Processor software. The article effectively demonstrates how generative grammars, pattern languages, and related extensions (e.g., pattern rules, homomorphisms, negative contexts, and remote contexts) can capture the complexity and flexibility of musical form, especially with regard to improvisation. The authors also illustrate the transition from an initial musicological focus in the earlier Bol Processor (BP1) to the more generalized and compositional orientation of BP2, highlighting the system’s role in computer-assisted composition. Throughout, there is a strong emphasis on integrating ethnomusicological insights, computational methods, and practical software design.

Strengths

Comprehensive Description of Formal Methods

The paper provides a thorough explanation of how different grammar types — finite-state automata, context-free grammars, type-0 grammars — are relevant for describing musical forms. This is complemented by examples that clarify the theoretical concepts (e.g., the use of pattern rules to handle repeated sections or “voiced/unvoiced” transformations). These sections effectively detail both traditional and innovative model expansions.

Contextual Application in Ethnomusicology

By applying generative models to North Indian tabla’s “theme-and-variations” structures (qa‘idas), the paper demonstrates how such computational approaches can illuminate facets of an orally transmitted musical tradition. The depth of collaboration with expert musicians and the iterative process of evaluation is indicative of a conscientious ethnographic methodology.

Transition to a Generalized Composition Environment

BP2 goes beyond traditional ethnomusicological applications and can be employed as a computer-aided composition tool for broader musical contexts. The concept of “sound-objects” and flexible time-handling approaches (striated vs. smooth time, time pivots, constraint satisfaction, etc.) make the workflow appealing to composers and researchers interested in generative music systems.

Balanced Presentation of Achievements and Challenges

The paper addresses problems of knowledge transfer, data representation, and the complexity of rule-based systems, providing valuable lessons for future computational musicology projects. It also describes how the QAVAID subsystem attempts to automate knowledge acquisition and inference, thereby removing some of the bottlenecks encountered when manually constructing grammars.

Extensive Reference to Prior Research

A substantial and well-organized reference section places the work in conversation with relevant literature from formal language theory, musicological studies, and AI-based composition systems. The inclusion of theoretical sources (e.g., citations to Chomsky’s hierarchy) and practical references (e.g., manual or shareware availability) underscores the paper’s academic depth and real-world applicability.

Overall Appraisal

This study offers a detailed and innovative account of how formal generative tools can be harnessed to describe, analyze, and produce complex musical structures. By bridging theoretical computer science, ethnomusicological fieldwork, and compositional practice, the manuscript opens possibilities for deeper exploration of music’s syntactic features and creative applications in contemporary composition. Its emphasis on practical software development, evaluated in tandem with expert critiques, positions it as a noteworthy resource for researchers and practitioners at the intersection of musicology, AI, and digital arts.

Download this paper

Skip to PDF content

Bol Processor grammars

James Kippen & Bernard Bel

In Mira Balaban, Otto Laske et Kemal Ebcioglu (eds.) Understanding Music with AI, American Association for Artificial Intelligence Press, Menlo Park CA (1992): 366-400.

Abstract

Bol Processor grammars are an extension of unrestricted generative grammars allowing a simple representation of string "patterns", here taken to mean repetitions and homomorphic transformations. These have been successfully applied to the simulation of improvisatory techniques in traditional drum music, using a production-rule system called "Bol Processor BP1". The basic concepts and parsing techniques of BP1 are presented.

A new version of Bol Processor, namely "BP2", has been designed to serve as a aid to rule-based composition in contemporary music. Extensions of the syntactic model, such as metavariables, remote contexts, substitutions and programmed grammars, are briefly introduced.

Excerpts of an AI review of this paper (Academia, June 2025)

Overview and Summary

The authors propose a formalism, called "Bol Processor grammars," designed to capture and simulate performance and improvisatory behaviors in traditional drum music—particularly North Indian tabla. This work presents a detailed account of how their proposed grammar-based system (BP1 and subsequently BP2) manages musical elements such as generative rules, parsing procedures, and higher-order transformations. The authors draw on concepts from formal language theory, specifically incorporating string rewriting, generative grammars of varying types (from context-free to unrestricted), and pattern languages.

The monograph not only discusses theoretical frameworks but also provides implementations and examples relevant to composition, improvisation, and ethnomusicological analysis. By combining standard grammars with additional features (e.g., pattern rules, negative contexts, remote contexts, substitutions, homomorphisms, and a sophisticated weighting mechanism), the Bol Processor aims to model creative aspects of improvisatory traditions.

Contribution and Significance

  • The paper bridges theoretical computer science (rewriting systems, generative grammars) with ethnomusicological inquiry. This interdisciplinary approach shows how language models can adapt to musical performance traditions, especially where oral transmission prevails.
  • The authors introduce extensions to classical Chomsky hierarchies by incorporating string pattern languages and homomorphisms specifically tailored for music composition and analysis. This advancement is especially valuable to those researching computational musicology or algorithmic composition.
  • By providing practical implementation details and guidelines (e.g., subgrammars, weighting rules, context-sensitive substitutions), the study conveys a clear path for others looking to model or simulate improvisational processes.

Strengths

  • Clarity of Theoretical Underpinnings: The text carefully explains the fundamentals of generative grammars and pattern languages, ensuring that readers unfamiliar with formal language theory can still follow the rationale behind the Bol Processor model.
  • Comprehensive Examples: The inclusion of worked-through grammar listings, detailed parsing traces, and real-world musical segments highlights an applied perspective. Readers can see exactly how the rules operate on concrete musical materials.
  • Interdisciplinary Integration: The manuscript thoughtfully weaves together computational linguistics, ethnomusicology, and composition, offering a unique perspective to each discipline.
  • Generative and Analytical Capacities: Emphasizing both the generation of new musical variations and the parsing of existing performances demonstrates the system’s two-fold utility: it supports creative composition while providing a framework for empirical analysis.

Areas for Further Development

  • Handling of Larger-Scale Form: While the paper addresses theme-variation structures, the methodology could be extended to more extensive global forms or multi-sectional pieces. More elaboration on how the grammar might manage nested forms or very long structures would strengthen the approach.
  • Quantitative Evaluation: The text provides evidence of successful modeling but could benefit from additional discussion of how correctness or “musical plausibility” is systematically tested, beyond anecdotal or interactive sessions with experts.
  • Comparisons with Other Systems: A more in-depth comparison with existing computational music systems that also employ generative grammars (e.g., purely context-free or Markov-based approaches) might deepen an understanding of the Bol Processor’s unique contributions.

Readability and Presentation

  • The writing is clear and consistently structured, especially around the discrete sections (introduction, pattern rules, parsing, and advanced features in BP2). Diagrams and grammar listings are helpful, though further clarifying annotations in some figures could assist readers less familiar with formal notation.
  • The appendices excel in presenting extended examples and step-by-step parses, adding transparency to the methods. This style of presentation ensures reproducibility and offers insights into how to adapt or modify the system for other musical styles.

Potential Impact on the Field

The system has apparent implications not just for tabla and other percussion traditions but for any domain where complex variations can be expressed in a rule-based manner. Likewise, composers working with algorithmic or computer-aided composition may discover a robust set of techniques for shaping variation, texture, and form. Researchers in ethnomusicology might find new analytical tools for uncovering systematic elements in improvisational practices.

Overall, the work stands as a thorough exploration of generative and parsing approaches tailored to music, illustrating how computational models can deepen understanding of both fixed and improvised musical structures.

Download this paper

Skip to PDF content

« Translate