pith. sign in

arxiv: 2604.21822 · v1 · submitted 2026-04-23 · 💻 cs.SD

Beyond Rules: Towards Basso Continuo Personal Style Identification

Adam \v{S}tefunko , Jan Haji\v{c} jr This is my paper

Pith reviewed 2026-05-08 13:28 UTC · model grok-4.3

classification 💻 cs.SD
keywords basso continuopersonal style identificationplayer classificationgriffs representationsupport vector machineshistorical performance practicecomputational musicologyimprovised accompaniment
0
0 comments X

The pith

Basso continuo players can be identified from their pitch choices in performance using structured representations and machine learning.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether individual performers leave detectable traces in their basso continuo realizations beyond the shared historical rules. It converts performance data into a pitch-focused representation called griffs and trains support vector machines to classify which player produced each realization. Successful classification indicates that personal stylistic preferences exist alongside the theoretical constraints. This shifts study of the genre from rule-based analysis to empirical examination of actual playing practices.

Core claim

Using the ACoRD dataset and its alignment of realizations to scores, the authors represent each performance as griffs—a structured encoding of the pitch content—and apply support vector machines to classify the performer. The models succeed at identifying individual players, and the paper examines which features of the griffs capture the distinctive elements of each player's style.

What carries the argument

Griffs, a historically informed structured representation of basso continuo performance pitch content, paired with support vector machine classification to detect player-specific patterns.

If this is right

  • Personal stylistic signatures can be quantified even in highly rule-governed improvised genres.
  • The same griffs representation can be used to compare stylistic traits across multiple performers.
  • Empirical methods can supplement the study of historical treatises by measuring how performers actually deviate from or interpret the rules.
  • Similar classification pipelines could be applied to other keyboard or ensemble improvisation traditions with aligned performance data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If griffs-based classification works reliably, it could support tools that help students match their realizations to historical exemplars of a chosen style.
  • The approach might extend to automatic generation systems that imitate specific performers rather than generic rules.
  • Longer-term, collections of classified realizations could reveal how individual styles evolve or cluster within broader historical schools of playing.

Load-bearing premise

That accurate classification from griffs captures genuine differences in personal style rather than effects from recording conditions, instruments, or choices in how the data were prepared.

What would settle it

Re-run the same SVM experiments on a new set of performances where the same players realize identical bass lines under matched recording and instrument conditions; classification accuracy should drop sharply if the original result was driven by non-stylistic factors.

Figures

Figures reproduced from arXiv: 2604.21822 by Adam \v{S}tefunko, Jan Haji\v{c} jr.

Figure 1
Figure 1. Figure 1: Griffs are extracted from aligned performances (a). A griff is a sequence of vectors formed by notes appearing together within a time window (b), transformed from pitches to intervals (c), and encoded as strings (d) that can form n-grams (e). symbol, which in our case is a hashtag. In the opposite direction, we can get rid of the structural aspect of griffs and simply convert the pitch of every performance… view at source ↗
Figure 2
Figure 2. Figure 2: Accuracy distributions of different score notes using segments of different size aggregated over the whole view at source ↗
Figure 3
Figure 3. Figure 3: Score note mean accuracies using segments of different lengths. Musical staves below each plot represent view at source ↗
Figure 4
Figure 4. Figure 4: Griff type distributions among different players for a high and a low accuracy note of Score 003. Each bar on the x-axis represents a different griff type. 6 Conclusion Basso continuo, as an improvisatory practice, is rooted in historical stylistic traditions which provide a constrained framework in which basso continuo realizations are improvised. However, basso continuo also gives its practitioner some f… view at source ↗
read the original abstract

A central part of the contemporary Historically Informed Practice movement is basso continuo, an improvised accompaniment genre with its traditions originating in the baroque era and actively practiced by many keyboard players nowadays. Although computational musicology has studied the theoretical foundations of basso continuo expressed by harmonic and voice-leading rules and constraints, characteristics of basso continuo as an active performing art have been largely overlooked mostly due to a lack of suitable performance data that could be empirically analyzed. This has changed with the introduction of The Aligned Continuo Realization Dataset (ACoRD) and the basso continuo realization-to-score alignment. Basso continuo playing is shaped by stylistic traditions coming from historical treatises, but it also may provide space for showcasing individual performance styles of its practitioners. In this paper, we attempt to explore the question of the presence of personal styles in the basso continuo realizations of players in the ACoRD dataset. We use a historically informed structured representation of basso continuo performance pitch content called griffs and Support Vector Machines to see whether it is possible to classify players based on their performances. The results show that we can identify players from their performances. In addition to the player classification problem, we discuss the elements that make up the individual styles of the players.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper introduces 'griffs' as a historically informed structured representation of basso continuo performance pitch content from the ACoRD dataset and applies Support Vector Machines to classify individual players based on their realizations. It claims that this enables player identification from performances and discusses the elements constituting individual performance styles beyond theoretical rules.

Significance. If the classification results prove robust under proper controls and validation, the work would be significant for computational musicology and historically informed performance studies. It provides an empirical bridge between rule-based theoretical models of basso continuo and actual performance data, potentially enabling new analyses of personal stylistic variation in improvisation. The choice of a domain-informed representation (griffs) rather than generic features is a clear strength that enhances interpretability.

major comments (2)
  1. [Results] Results section: The central claim that 'we can identify players from their performances' is not supported by any reported accuracy metrics, cross-validation procedure, baseline comparisons, statistical significance tests, or controls for potential confounders (e.g., instrument differences, recording conditions, or dataset biases). Without these, the claim that griffs + SVM captures genuine personal style rather than artifacts cannot be evaluated.
  2. [Method] Method section: The definition and construction of 'griffs' is introduced as the key representation, but the manuscript provides insufficient detail on its exact feature extraction process, dimensionality, invariance properties, or how it differs from standard pitch-class or harmonic representations. This is load-bearing because the classification success is attributed specifically to this representation.
minor comments (3)
  1. [Abstract] Abstract: The statement 'the results show that we can identify players' should be accompanied by at least the headline accuracy or F1 score to allow readers to gauge the strength of the finding immediately.
  2. [Discussion] Discussion: The qualitative discussion of stylistic elements would benefit from explicit linkage back to specific griff features or SVM decision boundaries to make the interpretation more falsifiable.
  3. [Introduction] Notation: The term 'griffs' is used without an initial formal definition or reference to its etymology/prior usage in continuo literature; a brief clarifying sentence or footnote would improve accessibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We are grateful to the referee for their thorough review and valuable suggestions. We provide point-by-point responses to the major comments and will incorporate revisions to address the concerns raised.

read point-by-point responses

  1. Referee: [Results] Results section: The central claim that 'we can identify players from their performances' is not supported by any reported accuracy metrics, cross-validation procedure, baseline comparisons, statistical significance tests, or controls for potential confounders (e.g., instrument differences, recording conditions, or dataset biases). Without these, the claim that griffs + SVM captures genuine personal style rather than artifacts cannot be evaluated.

    Authors: We agree that the Results section requires more quantitative support for the classification claim. The manuscript will be revised to include specific accuracy metrics from the SVM classifications, a detailed description of the cross-validation procedure employed, comparisons against baseline methods (such as random guessing or alternative feature representations), statistical significance testing (e.g., p-values or confidence intervals), and explicit discussion of controls for confounders including instrument variations and recording conditions. These additions will substantiate that the griffs-based approach identifies personal styles beyond potential artifacts. revision: yes

  2. Referee: [Method] Method section: The definition and construction of 'griffs' is introduced as the key representation, but the manuscript provides insufficient detail on its exact feature extraction process, dimensionality, invariance properties, or how it differs from standard pitch-class or harmonic representations. This is load-bearing because the classification success is attributed specifically to this representation.

    Authors: We acknowledge the need for greater detail on the griffs representation. In the revised manuscript, the Method section will be expanded to include: the precise algorithm for extracting griffs from the aligned performance data, the resulting feature dimensionality, properties regarding invariance (such as to key transposition or rhythmic variations), and a comparative analysis showing how griffs differ from and extend beyond standard pitch-class profiles or harmonic rule-based representations. This will clarify why griffs are particularly suited for capturing personal stylistic elements in basso continuo performance. revision: yes

Circularity Check

0 steps flagged

No significant circularity in empirical classification pipeline

full rationale

The paper presents a standard supervised classification task: it extracts 'griffs' (a historically informed pitch-content representation) from the ACoRD dataset and applies SVM to identify players. No derivation chain, equations, or self-citations reduce the central claim to its own inputs by construction. The reported ability to classify players is an empirical outcome of the chosen features and classifier on external performance data, not a fitted parameter renamed as prediction or a self-referential definition. The approach remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim depends on the griffs representation validly encoding stylistic information and on the assumption that player identity is the dominant source of variation in the realizations.

axioms (2)
  • domain assumption The griffs representation adequately captures pitch content relevant to personal style in basso continuo.
    Used as the input feature for SVM classification of players.
  • domain assumption Differences across realizations are primarily attributable to individual performer style rather than other factors.
    Required to interpret classification success as evidence of personal styles.
invented entities (1)
  • griffs no independent evidence
    purpose: Structured representation of basso continuo performance pitch content for analysis
    Introduced or applied as a historically informed encoding of the performance data.

pith-pipeline@v0.9.0 · 5518 in / 1364 out tokens · 80195 ms · 2026-05-08T13:28:55.600901+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

  1. [1]

    On every note a griff: Looking for a useful representation of basso continuo performance style

    Adam Štefunko, Carlos Eduardo Cancino-Chacón, and Jan Haji ˇc jr. On every note a griff: Looking for a useful representation of basso continuo performance style. InMusic Encoding Conference 2026,

    work page 2026

  2. [2]

    Automatic stylistic composition of bach chorales with deep lstm

    Feynman T Liang, Mark Gotham, Matthew Johnson, and Jamie Shotton. Automatic stylistic composition of bach chorales with deep lstm. InISMIR, volume 2017, page 18th,

    work page 2017

  3. [3]

    Lin Zhu and Wenjuan Zhang

    URLhttps://doi.org/10.5281/zenodo.16948235. Lin Zhu and Wenjuan Zhang. Research on polyphonic music generation algorithm based on gpt large model.IEEE Access, 13:129638–129647,

    work page doi:10.5281/zenodo.16948235

  4. [4]

    Cheng-Zhi Anna Huang, Tim Cooijmans, Adam Roberts, Aaron C

    doi:10.1109/ACCESS.2025.3588847. Cheng-Zhi Anna Huang, Tim Cooijmans, Adam Roberts, Aaron C. Courville, and Douglas Eck. Counterpoint by convolution. In Sally Jo Cunningham, Zhiyao Duan, Xiao Hu, and Douglas Turnbull, editors,Proceedings of the 18th International Society for Music Information Retrieval Conference, ISMIR 2017, Suzhou, China, October 23-27,...

    work page doi:10.1109/access.2025.3588847 2025

  5. [5]

    URL https://ismir2017.smcnus.org/wp-content/uploads/2017/10/187_Paper. pdf. Adam Štefunko, Suhit Chiruthapudi, Jan Haji ˇc jr., and Carlos Eduardo Cancino-Chacón. Basso continuo goes digital: Collecting and aligning a symbolic dataset of continuo performance, September

    work page 2017

  6. [6]

    Suhit Chiruthapudi, Adam Štefunko, Jan Haji ˇc jr., and Carlos Eduardo Cancino-Chacón

    URL https: //doi.org/10.5281/zenodo.16946799. Suhit Chiruthapudi, Adam Štefunko, Jan Haji ˇc jr., and Carlos Eduardo Cancino-Chacón. Challenges in basso con- tinuo performance-to-score alignment. InProceedings of the 17th International Symposium on Computer Music Multidisciplinary Research, pages 924–936. Zenodo, November

    work page doi:10.5281/zenodo.16946799

  7. [7]

    URL https://doi.org/10.5281/zenodo.17502511

    doi:10.5281/zenodo.17502511. URL https://doi.org/10.5281/zenodo.17502511. Georg Muffat. Regulae concentuum partiturae,

    work page doi:10.5281/zenodo.17502511