Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

AnalysisGNN claims that a single graph neural network, trained by shuffling mini-batches across heterogeneous corpora and fusing task logits, can match single-corpus models while staying resilient to domain shift.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 23:18 UTC pith:KLTXPSBW

load-bearing objection Useful multi-task GNN with a valuable corpus combination, but the 'domain-shift resilience' headline is not backed by the experiments—the joint model is only ever tested on training corpora. the 4 major comments →

arxiv 2509.06654 v1 pith:KLTXPSBW submitted 2025-09-08 cs.SD cs.AI

AnalysisGNN: Unified Music Analysis with Graph Neural Networks

Emmanouil Karystinaios , Johannes Hentschel , Markus Neuwirth , Gerhard Widmer This is my paper
classification cs.SD cs.AI
keywords music analysisgraph neural networksmulti-task learningharmonic analysiscadence detectionRoman numeral analysisnon-chord-tone predictiondomain shift
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

AnalysisGNN sets out to solve a fragmentation problem in computational music analysis: harmony, cadence, phrase, and meter are usually modeled by separate systems trained on separate datasets, each with its own annotation conventions. The paper argues a single graph neural network can be trained on all corpora at once, treating every score as a graph of notes and predicting twenty-odd note-level properties through shared features plus cross-task attention. Trained this way, the model is not the best on any single benchmark, but it is competitive, and it holds up when evaluated on a corpus it did not specialize for—where specialized models lose accuracy. The specific mechanism the paper adds is a non-chord-tone head that flags passing and non-functional notes and, at inference, lets only chord tones through to the task heads. If the claim holds, unified training could replace per-dataset pipelines and make cross-corpus analysis and benchmarking practical.

Core claim

The central claim is that heterogeneous annotation schemes can be reconciled not by harmonizing labels first, but by training a shared encoder with a multi-task objective that samples across all corpora each step, weights each task's loss by a learned uncertainty, and lets task-specific classifiers exchange information through an attention-based logit-fusion layer. On the combined corpus the model reaches 0.530 CSR on the smaller Roman-numeral set and 0.516 on the larger, internally consistent one, with cadence F1 of 0.558 on the larger corpus—compared with single-corpus baselines that are better on their home corpus but collapse elsewhere. The joint model's larger-corpus Roman-numeral score

What carries the argument

The load-bearing mechanism is logit fusion between task-specific classifiers, layered on a hybrid graph-convolution/GRU encoder. Each task head emits raw logits, projects them into a common space, and a transformer-style self-attention layer lets every head attend to the others before the final prediction; this turns separate analysis tasks into mutually corrective signals. The second part is the non-chord-tone prediction head, which at inference gates predictions to notes classified as chord tones. The third is the training schedule: mini-batches sampled across all tasks with an uncertainty-weighted loss, so no corpus dominates and label conflicts are balanced.

Load-bearing premise

The paper's unified label mapping—Roman numerals simplified to root and quality, local keys, and inversion—is consistent enough across corpora that joint training transfers knowledge instead of averaging conflicting annotations; if that mapping loses distinctions real annotators make, the reported domain-shift resilience is partly an artifact of coarser labels.

What would settle it

Train AnalysisGNN on the combined corpora but evaluate it on a held-out corpus whose Roman-numeral annotations use a fuller expressive vocabulary (including details the simplification drops); if performance collapses relative to a single-corpus model, the claimed resilience is an artifact of label simplification. A controlled ablation that turns off the non-chord-tone gate at inference while keeping every other component identical would also settle whether the gate contributes any of the cadence or phrase gains.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • A single AnalysisGNN model can be applied to any corpus with any subset of the supported annotations, including pieces with missing or partial labels, because invalid labels are masked rather than discarded.
  • Cross-task training transfers positively: knowing which notes are functional improves cadence detection, and harmonic context helps identify chord tones, so future systems can add auxiliary tasks as a form of self-supervision.
  • The resilience numbers imply that annotation-schema differences need not block joint training; benchmarks could be built by pooling existing corpora instead of waiting for one perfectly consistent dataset.
  • Non-chord-tone filtering at inference gives a cheap way to reduce false positives at cadence boundaries and focus phrase and section predictions on structurally relevant notes, which could carry over to higher-level form analysis.
  • Because the model predicts 20-plus note-level properties from one shared representation, analysis tools can present harmony, cadence, and phrasing as one coherent output rather than as separate pipelines.

Where Pith is reading between the lines

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

  • Editorial inference: the paper's most striking result—positive cadence transfer on the larger corpus (F1 rising from 0.479 to 0.558)—suggests cadence detection may benefit more from shared harmonic context than from the cadence annotations themselves; testing cadence accuracy while ablating the harmonic heads would isolate that.
  • Editorial inference: the robustness to domain shift may partly trace to the simplified label vocabulary (root, quality, local key, inversion); if the unified mapping discards distinctions the original corpora make, then 'resilience' could be averaging over coarser labels. A test on corpora with full Roman-numeral vocabularies would settle how much robustness is real.
  • Editorial inference: the non-chord-tone gate is a test-time filtering mechanism that could be studied independently of multi-task training—for example, applying the trained gate to a frozen single-task model should quantify how much of the gain comes from filtering versus from shared representations.
  • Editorial inference: the same recipe—shuffle batches across corpora, weight losses by uncertainty, fuse task logits—is a generic template for any domain with overlapping but inconsistent annotation layers, such as multi-annotator medical labels or multilingual parsing; the graph-specific details are secondary.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. AnalysisGNN proposes a unified graph-neural-network framework for multiple symbolic music analysis tasks, trained jointly on the Distant Listening Corpus (DLC), the AugmentedNet dataset, and three cadence-annotated corpora. The model combines a HybridGNN encoder with a data-shuffling training strategy, a learnable weighted multi-task cross-entropy loss, cross-task logit fusion via self-attention, and a Non-Chord-Tone (NCT) prediction head that gates out predicted non-chord-tones at inference. Experiments compare AnalysisGNN against single-corpus baselines and report configuration ablations. The paper claims performance comparable to specialist single-task systems and increased resilience to domain shifts and annotation inconsistencies across heterogeneous corpora.

Significance. If the empirical claims were fully supported, this would be a useful contribution to computational music analysis: the paper releases source code, assembles a large heterogeneous corpus collection, introduces new note-level analytical tasks, and proposes a plausible mechanism for joint training across inconsistent annotation standards. The architectural ideas—data shuffling, weighted multi-task loss, logit fusion, and NCT-based filtering—are reasonable and worth exploring. However, the central claim of domain-shift resilience is not actually tested, and several load-bearing comparisons rest on small, single-run performance differences without uncertainty estimates. The framework is promising, but the current evidence is insufficient for the strength of the abstract and conclusion claims.

major comments (4)
  1. [Section 5.1, Table 3] The domain-shift resilience claim is not tested. In Table 3, the 'All corpora (combined)' row trains on all three corpora and evaluates on test splits of those same corpora; these are in-domain evaluations, not held-out domain-shift tests. The single-corpus rows show that single-corpus models degrade when evaluated on another corpus, but they do not show that AnalysisGNN generalizes to an unseen corpus. The RNBert comparison (RNBert trained on AugNet evaluated on DLC, CSR 0.301, versus AnalysisGNN CSR 0.516) is confounded because AnalysisGNN was trained on DLC itself. To support the claim, the authors need to train on a subset of corpora and evaluate on a held-out corpus, or explicitly reframe the claim as 'multi-corpus training trade-off' rather than domain-shift resilience.
  2. [Section 5.2, Tables 2-4] All results are reported as single runs without error bars or statistical significance. Several differences that drive the narrative are small—for example, DLC cadence F1 changes from 0.556 (DLC-only) to 0.558 (all corpora), and phrase DLC F1 changes from 0.752 (without logit fusion) to 0.742 (full model). These differences may be within run-to-run noise. Since the paper's central claims about positive transfer, minor performance drops, and component contributions depend on numerical comparisons, multiple seeds with standard deviations (or significance tests) are necessary.
  3. [Section 3.2, Table 4] The contribution of the NCT gating mechanism is not ablated. The paper claims that excluding predicted non-chord-tones at inference improves the consistency of label signals, but the configuration study only reports 'w/o Aux-Tasks', which removes auxiliary tasks as a group, not specifically the NCT head or the inference-time gate. A controlled comparison with and without the NCT gate at inference is needed to support the abstract's claim about the NCT module and the conclusion's claim about 'cleaner label signals and faster inference'.
  4. [Section 4.1, Table 3] The label unification step—Roman numerals simplified to root/quality, local keys, and inversions—is not validated. Table 3 shows that DLC-only Roman numeral CSR drops from 0.563 (DLC-only) to 0.516 (all corpora). If the unification is lossy, this drop may reflect conflicting annotation schemas being averaged rather than genuine robustness. The authors should quantify how often the simplified labels disagree across the two datasets, and discuss whether the observed drop is acceptable under their unification or is an artifact of schema tension.
minor comments (5)
  1. [Section 5.1] The text refers to 'Figure 3' for single-corpus versus joint-corpus comparisons, but no Figure 3 appears in the manuscript. This is either a missing figure or a broken cross-reference.
  2. [Section 4.2] Typo: 'DCL' should be 'DLC' in the first paragraph.
  3. [Section 6] Typo: 'comparativeaccuracy' should be 'comparative accuracy'.
  4. [Table 1] The 'labels*' column is defined only implicitly; clarify in the caption what 'original files before the expansion' means and why only DLC has this column.
  5. [Section 4.3] The paper states the model predicts 20 properties, but experimental results report only cadence, Roman numeral, phrase, pedal, metrical, and section. Consider providing a full breakdown in a supplementary table, or clearly noting that additional tasks are not evaluated.

Circularity Check

0 steps flagged

No circular derivation; domain-shift claim is untested but not circular.

full rationale

The paper contains no mathematical derivation whose conclusion is equivalent to its assumptions. All headline numbers (RN CSR, cadence F1, phrase F1) are computed on held-out test splits and compared with external baselines (AugmentedNet, ChordGNN, RNBert), so the main empirical results are not constructed from the fitted parameters. The learnable loss weights sigma_t in Eq. (1) are trained on training folds and do not themselves define any reported accuracy. The cited earlier work by the same authors ([8,9,10]) supplies the note-graph representation, ChordGNN encoder, GraphMuse preprocessing, and neighbor sampling; these are reusable building blocks and are not invoked as an external proof of the central multi-task claim, so they are self-citations but not load-bearing. The one real concern is that the abstract's 'increased resilience to domain shifts' is not actually tested: Table 3's 'All corpora (combined)' row evaluates on test sets whose corpora (Cadence, AugNet, DLC) are all present in the combined training set, and the paper never trains on a subset and evaluates on a held-out corpus. This is an experimental-design/validity gap, not a circular reduction: the in-domain test numbers are not definitionally equal to the training loss or to any fitted quantity, and a properly held-out corpus experiment could in principle confirm or refute the claim. Per the review rule, this missing support is flagged and weighed, but it does not make the derivation circular. Score 2 reflects the presence of non-load-bearing self-citations; the central content is otherwise self-contained.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 1 invented entities

The central claims rest on standard neural-network machinery, a graph-based score representation, and the compatibility of three annotation schemes. The NCT gating introduces an additional ad hoc assumption with no isolated validation.

free parameters (5)
  • Task loss scales sigma_t (Eq. 1) = not reported
    Learnable per-task weights in the multi-task loss, optimized on training data by gradient descent; values not reported in the paper.
  • Hidden size / output size = 256 / 128
    Hand-set architecture hyperparameters in Section 5.3; define model capacity.
  • Optimizer hyperparameters = lr=0.005, wd=0.0005, warmup=500, batch=250, subgraph=500
    Hand-set in Section 5.3; not tuned per task or dataset.
  • Projection dimension d for logit fusion = unknown
    Dimension of the common space into which each task's logits are projected before self-attention (Section 3.2); not specified in the paper.
  • Transposition augmentation multiplier = approximately 10x
    Data augmentation strategy in Section 4.2 chosen to roughly decuple training data; hand-chosen.
axioms (5)
  • domain assumption A musical score can be represented as a graph of note nodes with temporal edges preserving analysis-relevant structure
    Invoked in Section 3.1, building on [7,8]; if the graph loses voice-leading or long-range context, the model's inputs are insufficient.
  • domain assumption Heterogeneous annotation schemas can be merged into a single label space without systematic conflict
    Used in Section 4.1 to combine DLC, AugmentedNet, and cadence corpora; contradicted in part by Table 3's in-domain drops.
  • domain assumption Non-chord-tone status is learnable from note-level features and is a valid proxy for functional importance
    Underpins the NCT module in Section 3.2; no accuracy evaluation of the NCT head is reported.
  • ad hoc to paper Gating out predicted non-chord-tones at inference improves downstream task predictions
    Adopted in Section 3.2 ('pass only those identified as chord-tones to the task-specific heads') without an isolated ablation.
  • standard math Standard backpropagation, cross-entropy, and attention mechanisms behave as expected
    Implied by the training setup in Section 3.2; no formal guarantees required.
invented entities (1)
  • Non-Chord-Tone prediction module (NCT head and gate) no independent evidence
    purpose: To identify passing/non-functional notes and filter them from downstream task heads at inference, 'improving the consistency of label signals' (Abstract).
    The paper introduces this component but provides no external benchmark for NCT accuracy and no ablation isolating its effect; its claimed benefit is therefore not independently evidenced.

pith-pipeline@v1.3.0-alltime-deepseek · 8972 in / 16506 out tokens · 164678 ms · 2026-08-04T23:18:58.289517+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of AnalysisGNN: Unified Music Analysis with Graph Neural Networks." pith.science (2026). https://pith.science/paper/KLTXPSBW

@misc{pith2026250906654,
  author       = {Pith},
  title        = {Pith review of: AnalysisGNN: Unified Music Analysis with Graph Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KLTXPSBW}},
  note         = {Machine review of arXiv:2509.06654}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Recent years have seen a boom in computational approaches to music analysis, yet each one is typically tailored to a specific analytical domain. In this work, we introduce AnalysisGNN, a novel graph neural network framework that leverages a data-shuffling strategy with a custom weighted multi-task loss and logit fusion between task-specific classifiers to integrate heterogeneously annotated symbolic datasets for comprehensive score analysis. We further integrate a Non-Chord-Tone prediction module, which identifies and excludes passing and non-functional notes from all tasks, thereby improving the consistency of label signals. Experimental evaluations demonstrate that AnalysisGNN achieves performance comparable to traditional static-dataset approaches, while showing increased resilience to domain shifts and annotation inconsistencies across multiple heterogeneous corpora.

Figures

Figures reproduced from arXiv: 2509.06654 by Emmanouil Karystinaios, Gerhard Widmer, Johannes Hentschel, Markus Neuwirth.

Figure 1
Figure 1. Figure 1: Architecture overview: the input score graph is processed by stacked Hetero￾geneous GCN blocks in parallel with note-sequence features through an equally deep GRU; their shared embedding feeds 2-layer MLP task heads, whose outputs are then refined via a cross-task logit-fusion layer. 4 Corpora 4.1 Overview and Preprocessing Our work combines multiple datasets with diverse analytical annotations to train a … view at source ↗
Figure 2
Figure 2. Figure 2: AnalysisGNN predictions on the opening of Beethoven Sonata No.1, Op.2 Movement 1. The figure shows the score graph where red edges connect consecutive notes, blue edges link notes with the same onset, and green edges connect notes with onsets between the onset and offset of another. The model outputs harmonic analysis while highlighting passing notes. 5.3 Configuration All our models are trained on a singl… view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Tonnetz-Driven Graph Wedgelet for Harmonic Complexity Reduction in Music Scores

    cs.SD 2026-07 conditional novelty 5.5

    Adaptive graph wedgelets in a six-dimensional Tonnetz embedding compress piano subgraphs of vocal-pianistic scores into piecewise-constant, playable reconstructions that respect harmonic distance.

Reference graph

Works this paper leans on

24 extracted references · 21 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    Transactions of the International Society for Music Information Retrieval (TISMIR)2(1), 82–96 (2019)

    Allegraud, P., Bigo, L., Feisthauer, L., Giraud, M., Groult, R., Leguy, E., Levé, F.: Learning sonata form structure on mozart’s string quartets. Transactions of the International Society for Music Information Retrieval (TISMIR)2(1), 82–96 (2019)

    work page 2019

  2. [2]

    In: Proceedings of the International Society for Music Information Retrieval Conference (ISMIR) (2018)

    Bigo, L., Feisthauer, L., Giraud, M., Levé, F.: Relevance of musical features for cadence detection. In: Proceedings of the International Society for Music Information Retrieval Conference (ISMIR) (2018)

    work page 2018

  3. [3]

    In: International Joint Conference on Artificial Intelligence (IJCAI) (2023)

    Cosenza, E., Valenti, A., Bacciu, D.: Graph-based polyphonic multitrack music generation. In: International Joint Conference on Artificial Intelligence (IJCAI) (2023)

    work page 2023

  4. [4]

    Giraud,M.,Groult,R.,Leguy,E.,Levé,F.:Computationalfugueanalysis.Computer Music Journal39(2), 77–96 (2015)

    work page 2015

  5. [5]

    In: Proceedings of the European conference on computer vision (ECCV)

    Guo, M., Haque, A., Huang, D.A., Yeung, S., Fei-Fei, L.: Dynamic task prioritization for multitask learning. In: Proceedings of the European conference on computer vision (ECCV). pp. 270–287 (2018)

    work page 2018

  6. [6]

    https://doi.org/10.5281/zenodo.13844105,https://doi.org/ 10.5281/zenodo.13844105

    Hentschel, J., Rammos, Y., Neuwirth, M., Rohrmeier, M.: The Distant Listening Corpus (v3.1) (2025). https://doi.org/10.5281/zenodo.13844105,https://doi.org/ 10.5281/zenodo.13844105

    work page doi:10.5281/zenodo.13844105 2025

  7. [7]

    In: Proceedings of the 20th International Society for Music Information Retrieval Conference, ISMIR 2019, Delft, The Netherlands, November 4-8, 2019

    Jeong, D., Kwon, T., Kim, Y., Lee, K., Nam, J.: VirtuosoNet: A Hierarchical RNN-based System for Modeling Expressive Piano Performance. In: Proceedings of the 20th International Society for Music Information Retrieval Conference, ISMIR 2019, Delft, The Netherlands, November 4-8, 2019. pp. 908–915 (2019), http://archives.ismir.net/ismir2019/paper/000112.pdf

    work page 2019

  8. [8]

    In: Proceedings of the 23rd International Society for Music Information Retrieval Conference, ISMIR 2022, Bengaluru, India, December 4-8, 2022

    Karystinaios, E., Widmer, G.: Cadence Detection in Symbolic Classical Music using Graph Neural Networks.. In: Proceedings of the 23rd International Society for Music Information Retrieval Conference, ISMIR 2022, Bengaluru, India, December 4-8, 2022. pp. 917–924 (2022),https://archives.ismir.net/ismir2022/paper/ 000110.pdf

    work page 2022

  9. [9]

    In: Proceed- ings of the 24th International Society for Music Information Retrieval Con- ference, ISMIR 2023, Milan, Italy, November 5-9, 2023

    Karystinaios, E., Widmer, G.: Roman Numeral Analysis With Graph Neu- ral Networks: Onset-Wise Predictions From Note-Wise Features. In: Proceed- ings of the 24th International Society for Music Information Retrieval Con- ference, ISMIR 2023, Milan, Italy, November 5-9, 2023. pp. 597–604 (2023). https://doi.org/10.5281/ZENODO.10265357, https://doi.org/10.52...

    doi:10.5281/zenodo.10265357 2023

  10. [10]

    In: Proceedings of the 25th International Society for Music Information Re- trieval Conference, ISMIR 2024, San Francisco, California, USA and Online, Novem- ber 10-14, 2024

    Karystinaios, E., Widmer, G.: GraphMuse: A Library for Symbolic Music Graph Pro- cessing. In: Proceedings of the 25th International Society for Music Information Re- trieval Conference, ISMIR 2024, San Francisco, California, USA and Online, Novem- ber 10-14, 2024. pp. 651–658 (2024). https://doi.org/10.5281/ZENODO.14877421, https://doi.org/10.5281/zenodo.14877421

    doi:10.5281/zenodo.14877421 2024

  11. [11]

    arXiv preprint arXiv:1805.06334 (2018)

    Liebel, L., Körner, M.: Auxiliary tasks in multi-task learning. arXiv preprint arXiv:1805.06334 (2018)

    Pith/arXiv arXiv 2018

  12. [12]

    Hierarchical Symbolic Pop Music Generation with Graph Neural Networks

    Lim, W.Q., Liang, J., Zhang, H.: Hierarchical symbolic pop music generation with graph neural networks. arXiv preprint arXiv:2409.08155 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  13. [13]

    Advances in Neural Information Processing Systems36, 57226–57243 (2023)

    Liu, B., Feng, Y., Stone, P., Liu, Q.: Famo: Fast adaptive multitask optimization. Advances in Neural Information Processing Systems36, 57226–57243 (2023)

    work page 2023

  14. [14]

    In: Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)

    McLeod, A.P., Rohrmeier, M.A.: A modular system for the harmonic analysis of musical scores using a large vocabulary. In: Proceedings of the International Society for Music Information Retrieval Conference (ISMIR). pp. 435–442 (2021) 14 E. Karystinaios et al

    work page 2021

  15. [15]

    Transactions of the International Society for Music Information Retrieval (TISMIR)3(1), 42–54 (2020)

    Micchi, G., Gotham, M., Giraud, M.: Not all roads lead to rome: Pitch representa- tion and model architecture for automatic harmonic analysis. Transactions of the International Society for Music Information Retrieval (TISMIR)3(1), 42–54 (2020)

    work page 2020

  16. [16]

    In: Proceedings of the Interna- tional Society for Music Information Retrieval Conference (ISMIR)

    Micchi, G., Kosta, K., Medeot, G., Chanquion, P.: A deep learning method for enforcing coherence in automatic chord recognition. In: Proceedings of the Interna- tional Society for Music Information Retrieval Conference (ISMIR). pp. 443–451 (2021)

    work page 2021

  17. [17]

    ACL Anthology (2021)

    Mishra, S., Khashabi, D., Baral, C., Hajishirzi, H.: Cross-task generalization via natural language crowdsourcing instructions. ACL Anthology (2021)

    work page 2021

  18. [18]

    In: Proceedings of the 22nd International Society for Music Informa- tion Retrieval Conference, ISMIR

    Nápoles López, N., Gotham, M., Fujinaga, I.: AugmentedNet: A Roman Nu- meral Analysis Network with Synthetic Training Examples and Additional Tonal Tasks. In: Proceedings of the 22nd International Society for Music Informa- tion Retrieval Conference, ISMIR. pp. 404–411. ISMIR, Online (Nov 2021). https://doi.org/10.5281/zenodo.5624533,https://zenodo.org/re...

    work page arXiv 2021

  19. [19]

    Computer Music Journal28(3), 45–52 (2004), http://www.jstor.org/stable/ 3681508

    Raphael, C., Stoddard, J.: Functional harmonic analysis using probabilistic models. Computer Music Journal28(3), 45–52 (2004), http://www.jstor.org/stable/ 3681508

    work page 2004

  20. [20]

    In: Proceedings of the International Society for Music Information Retrieval Conference (ISMIR) (2024)

    Sailor, M.: RNBert: Fine-tuning a masked language model for roman numeral analysis. In: Proceedings of the International Society for Music Information Retrieval Conference (ISMIR) (2024)

    work page 2024

  21. [21]

    Journal of New Music Research47(1), 29–52 (2018)

    Sears, D.R., Pearce, M.T., Caplin, W.E., McAdams, S.: Simulating melodic and harmonic expectations for tonal cadences using probabilistic models. Journal of New Music Research47(1), 29–52 (2018)

    work page 2018

  22. [22]

    MIT Press, Cambridge, Mass

    Temperley, D.: The Cognition of Basic Musical Structures. MIT Press, Cambridge, Mass. (2001)

    work page 2001

  23. [23]

    Journal of New Music Research38(1), 3–18 (Mar 2009)

    Temperley, D.: A Unified Probabilistic Model for Polyphonic Music Analysis. Journal of New Music Research38(1), 3–18 (Mar 2009). https://doi.org/10.1080/09298210902928495, http://www.tandfonline.com/doi/ abs/10.1080/09298210902928495

    work page doi:10.1080/09298210902928495 2009

  24. [24]

    IEEE transactions on knowledge and data engineering34(12), 5586–5609 (2021)

    Zhang, Y., Yang, Q.: A survey on multi-task learning. IEEE transactions on knowledge and data engineering34(12), 5586–5609 (2021)

    work page 2021