Pith. sign in

REVIEW 3 major objections 6 minor 32 references

Semantic-Aware Interpretable Multimodal Music Auto-Tagging

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Interpretable banded regression beats transformer on MTG-Jamendo

desk verdict The explanation story—lambda magnitudes as group importance—is plausible but unvalidated, and the performance edge appears on only one of three datasets; still, the paper is honest, reproducible, and worth a serious referee. read the letter →

arxiv 2505.17233 v3 pith:2QI3V4EC submitted 2025-05-22 cs.LG cs.SDeess.AS

Andreas Patakis , Vassilis Lyberatos , Spyridon Kantarelis , Edmund Dervakos , Giorgos Stamou This is my paper
classification cs.LGcs.SDeess.AS
keywords musicauto-tagginginterpretabilitymultimodalfeaturesEM-BANDEDsemanticgroupingperceptuallyrics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper is trying to show that interpretability in music auto-tagging need not be sacrificed for predictive accuracy. It builds a pipeline of human-understandable perceptual features drawn from audio, harmony, and lyrics, groups them into semantically meaningful clusters, and fits a banded regression model that assigns each cluster its own shrinkage weight. The reported result is that this model matches or surpasses a transformer-based baseline on one dataset and offers explanations that non-expert listeners rate as more intuitive than those of a standard gradient-boosting model. The reason to care is that if this holds, users can trust and inspect why a tag was assigned, rather than treat the model as a black box.

What carries the argument

The load-bearing object is EM-BANDED, an expectation-maximization algorithm for banded regression: a linear model whose regression coefficients are partitioned into groups, each with its own variance hyperparameter $\lambda_j$ in a block-diagonal Gaussian prior. Estimated group-level $\lambda_j$ values shrink each group's weights differentially, and the paper reads the magnitude of $\lambda_j$ as that group's importance to the tag decision. The feature groupings themselves form the second component: three hand-designed semantic schemes (User-Friendly, Domain-Expert, and All-Features) that organize features from signal processing, deep neural nets, ontology-based harmony, and lyrics into interpretable clusters.

What would settle it

A concrete test: take the trained EM-BANDED model on Music4All with the User-Friendly grouping, compute lambda-based rankings, then ablate each group one at a time by replacing its features with random noise or permuting them, and measure the drop in accuracy. If removing the group with the lowest lambda causes a larger accuracy drop than removing the group with the highest lambda, the lambda-as-importance reading fails. Comparing lambda rankings to permutation-based group importance on the same data would settle whether the explanation is trustworthy.

Watch

Extended reading notes

Core claim

The paper claims that a banded ridge regression with group-wise hyperparameters, fitted by expectation maximization and trained on semantically clustered perceptual features from audio and lyrics, can reach state-of-the-art results on one benchmark (76.95% ROC-AUC on MTG-Jamendo, above the transformer baseline's 75.90%) and produce group-level importance scores that rank closer to human amateur musicians' rankings (absolute distance 8) than XGBoost's aggregated importances (distance 12). On the other two datasets the transformer baseline remains ahead, but the interpretable model stays competitive, and adding lyrics features improves most models.

Load-bearing premise

The assumption that a group's lambda magnitude directly indicates how relevant that group is to the model's predictions; the paper offers no experiment or derivation showing that higher lambda values correspond to larger actual predictive contributions.

Editorial extensions

If this is right

  • If the central claim holds, an interpretable model can be the best performer on a standard tag benchmark, not just a compromise between accuracy and transparency.
  • Group-level explanations can be validated directly against human judgments, offering a concrete test for interpretability claims in music tagging.
  • Lyrics features reliably add signal across most models, reinforcing that multimodality helps tagging.
  • Random grouping produces uniform importances, which the paper takes as evidence that the model's explanations reflect real structure rather than artifacts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The central interpretability claim rests implicitly on the assumption that $\lambda_j$ magnitudes track true predictive contribution; an ablation that removes or perturbs high-$\lambda$ groups and measures the performance drop would test that directly, but the paper does not run it.
  • The same banded-regression machinery could transfer to other domains with interpretable feature groups, such as medical or financial tabular data, where group-level explanations are desirable.
  • A larger human study with more participants and per-tag rankings would be needed to know whether the closeness to intuition generalizes beyond the five-group Music4All setting.
  • Combining EM-BANDED's group importances with post-hoc model-agnostic methods could bridge the gap between intrinsic and post-hoc interpretability.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes SAMAT, an interpretable music auto-tagging pipeline that combines audio- and lyric-derived perceptual features, groups them semantically (User-Friendly, Domain-Expert, All-Features), and applies the EM-BANDED banded regression model to obtain group-level importance scores. The main claims are that the method achieves competitive tagging performance on MTG-Jamendo (76.95% ROC-AUC, above MULTIMODAL and XGBoost) and that EM-BANDED's lambda-based group importances are intuitive, as supported by a 10-participant human evaluation. On Music4All and AudioSet, MULTIMODAL remains stronger, with EM-BANDED trailing by roughly 12 and 10 accuracy points, respectively. The core interpretability contribution rests on treating lambda magnitudes as direct indicators of group relevance.

Significance. If the interpretability claim were properly validated, the paper would make a useful contribution to interpretable music tagging: it introduces a multimodal feature set with semantic groupings and uses a banded regression model to obtain deterministic, group-level explanations without relying on post-hoc analysis of a black box. The comparisons to a transformer baseline on MTG-Jamendo are interesting, and the inclusion of lyric-derived features alongside audio features is a reasonable direction. However, the central validity of the lambda-as-importance mapping is not demonstrated, the cross-dataset performance claim is overstated, and the human evaluation is too small and informal to support the stated conclusion. These issues are load-bearing for the paper's central claims, so the manuscript needs substantive revision.

major comments (3)
  1. [Section 2.3, Eq. (3)] The entire interpretability contribution rests on the assertion that 'the magnitude of lambda serves as a direct indicator of that group's relevance to the model.' However, lambda_j is a prior variance hyperparameter, not an estimated effect size. The actual contribution of group j to the prediction is F_j beta_j, which depends on the fitted coefficients beta_j and on the covariance structure Omega_j. A group can have a large lambda (weak shrinkage) while its posterior weights are small, or vice versa. The manuscript provides no derivation, simulation, or ablation showing that the ordering of lambda values tracks the true predictive contribution of each group. Because groups differ in size and Omega_j is not specified, raw lambda values are not obviously comparable across groups. This is a load-bearing gap: if lambda does not track predictive relevance, the group-importance explanations are not trustworthy. Please add a synthetic-data experiment with known ground-truth group effects, or an ablation that perturbs or removes groups and compares the resulting performance change to the lambda ranking.
  2. [Section 3.2, Table 1] The abstract and text state that the method achieves 'competitive tagging performance,' but Table 1 shows that this is accurate only for MTG-Jamendo. On Music4All, MULTIMODAL achieves 57.34% versus 45.09% for the best EM-BANDED variant; on AudioSet, MULTIMODAL achieves 58.71% versus 48.15%. These are substantial gaps, not simply 'competitive.' The sentence in Section 3.2 that EM-BANDED 'remains competitive (45.09% and 48.15%)' should be revised or supported with error bars and statistical tests. As reported, the claim in the abstract 'achieves competitive tagging performance' is overstated.
  3. [Section 3.2, Table 2] The human evaluation used only 10 amateur musicians, reports no variance or significance test, and the difference in Absolute Difference (8 vs. 12) may well be within noise. The claim that EM-BANDED's learned group-level importances 'better align with users' intuition' is therefore not statistically supported. In addition, the random-group experiment only shows that importance becomes uniform when no structure is imposed; it does not test whether lambda ordering matches true group effects for the semantic groups. Please either strengthen the human evaluation with more participants and hypothesis tests, or soften the claim to a preliminary observation.
minor comments (6)
  1. [Section 2.2] The feature counts are inconsistent: User-Friendly uses 67 features, Domain-Expert uses 42, and All-Features is said to use all available features, but the total number of features described in Section 2.1 is not explicitly stated and the relationship between these counts is unclear. Please clarify how many features are used in each grouping and how they are selected.
  2. [References] References [21] and [25] are the same paper (Fuglsang et al., Imaging Neuroscience 2024) and should be consolidated into a single citation.
  3. [Throughout] There are repeated typographical inconsistencies, including 'XGB OOST' instead of 'XGBoost' (Sections 1, 3.2), 'M ULTIMODAL' (Section 3.2), and 'Music4all' (Section 3.2) alongside 'Music4All'. These should be uniformly corrected.
  4. [Table 1 caption] The caption notes that ROC-AUC is used for MTG-Jamendo and accuracy for Music4All and AudioSet. Because the metrics differ, readers should be cautioned that cross-dataset numeric comparisons are not meaningful; only within-dataset comparisons should be used.
  5. [Figure 2] The manuscript references Figure 2 for group importances, but the figure itself is not included in the text provided; please ensure all referenced figures are present and legible.
  6. [Contributions bullet list] The first contribution says 'Provided clear and certain global explanations' but 'certain' is vague and potentially misleading given the unvalidated lambda-to-importance mapping; please rephrase to describe the type of explanation actually delivered.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the core experimental results are benchmarked externally, and the interpretability mapping is an asserted assumption rather than a derivation that reduces to the target result.

full rationale

The paper's derivation chain is largely self-contained against external benchmarks. Tagging performance is measured on MTG-Jamendo, Music4All, and AudioSet against MULTIMODAL [7] and XGBoost, which are fixed baselines rather than fitted targets. The input features come from published signal-processing libraries (Essentia), pretrained components (Omnizart, Demucs, Whisper), and a perceptual DNN trained on [12]; none are functions of the auto-tagging labels. EM-BANDED is adopted from the non-overlapping authors of [25], and the empirical results in Table 1 are external. The one potentially circular-looking step is Section 2.3's statement that 'the magnitude of lambda serves as a direct indicator of that group's relevance to the model.' This is an asserted operationalization, not a derivation: if 'relevance' were defined as lambda, the claim would be tautological, but the paper does not derive the equation; it assumes it. The human evaluation and random-group experiment are independent controls. The self-citations ([7], [14], [20]) supply a baseline, a feature-extraction pipeline, and an ontology; they do not by themselves force the claimed result. Accordingly, there is no exhibited reduction of a prediction to an input, and the circularity score is low.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim depends on several imported and assumed components: the EM-BANDED shrinkage-parameter interpretation, the semantic validity of hand-made feature groups, and the accuracy of a chain of external tools. The paper does not isolate these with ablations or sensitivity analyses.

free parameters (3)
  • EM-BANDED inverse-gamma hyperparameters (eta, tau) = not reported
    These hyperparameters control group-level shrinkage and therefore the lambda_j values that are read as importances; the paper does not state their values or robustness.
  • Hand-designed semantic group assignments = User-Friendly: 67 features in 5 groups; Domain-Expert: 42 features in 5 groups
    The grouping is a free design choice rather than a learned or validated structure, and the paper only compares it to a random partition.
  • Number of groups per semantic category = 5, 5, 4
    The group counts are chosen by the authors; no sensitivity analysis is reported.
assumptions (5)
  • ad hoc to paper The magnitude of the EM-BANDED lambda coefficient is a direct indicator of feature-group relevance.
    Section 2.3 states 'the magnitude of lambda serves as a direct indicator of that group's relevance to the model' without formal derivation.
  • domain assumption The 10 amateur musicians' rankings represent typical user intuition.
    Section 3.2 reports a human evaluation with 10 amateurs and no demographics, recruitment details, or sample size justification.
  • domain assumption The hand-crafted User-Friendly and Domain-Expert groupings reflect meaningful semantic categories.
    Section 2.2 says groups are produced by heuristic categorization by music experts; no inter-rater agreement or independent validation is provided.
  • domain assumption Automated extraction tools (Omnizart, Whisper, Demucs, Essentia) produce sufficiently accurate features.
    Section 2.1 relies on these tools for chord recognition, lyric transcription, source separation, and acoustic features; errors propagate to the regression.
  • domain assumption Perceptual DNN features trained on the dataset of [12] generalize to these tagging datasets.
    Section 2.1 transfers a VGG-style model to extract 7 perceptual attributes; no validation on target data is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Semantic-Aware Interpretable Multimodal Music Auto-Tagging." pith.science (2026). https://pith.science/paper/2QI3V4EC

@misc{pith2026250517233,
  author       = {Pith},
  title        = {Pith review of: Semantic-Aware Interpretable Multimodal Music Auto-Tagging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2QI3V4EC}},
  note         = {Machine review of arXiv:2505.17233}
}
read the original abstract

Music auto-tagging is essential for organizing and discovering music in extensive digital libraries. While foundation models achieve exceptional performance in this domain, their outputs often lack interpretability, limiting trust and usability for researchers and end-users alike. In this work, we present an interpretable framework for music auto-tagging that leverages groups of musically meaningful multimodal features, derived from signal processing, deep learning, ontology engineering, and natural language processing. To enhance interpretability, we cluster features semantically and employ an expectation maximization algorithm, assigning distinct weights to each group based on its contribution to the tagging process. Our method achieves competitive tagging performance while offering a deeper understanding of the decision-making process, paving the way for more transparent and user-centric music tagging systems.

Figures

Figures reproduced from arXiv: 2505.17233 by the authors.

Figure 1
Figure 1. Overview of the pipeline of our methodology. 2.1. Extracting Perceptual Multimodal Features Building upon the pipeline proposed in [14], we extracted per￾ceptual features using four distinct techniques: deep learning models, ontology engineering, signal processing, and natural language processing. We used symbolic knowledge to extract harmonic features based on music theory and harmony, building semantic graph datab… view at source ↗
Figure 2
Figure 2. Group importances for Music4All for User-Friendly and Domain-Expert approaches. BANDED achieves the highest performance (76.95%) on MTG￾Jamendo, surpassing both MULTIMODAL (75.90%) and XG￾BOOST (74.64%). On Music4All and AudioSet, MULTI￾MODAL performs best (57.34% and 58.71% respectively), while EM-BANDED remains competitive (45.09% and 48.15%). Lyri￾cal features consistently improve performance in almost all models… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 24 canonical work pages

  1. [1]

    This task involves as- signing semantic labels—such as genres, instruments, moods, and production qualities—to audio tracks

    Introduction In the era of expansive digital music repositories, automatic mu- sic auto-tagging has become essential for improving discover- ability, personalization, and organization. This task involves as- signing semantic labels—such as genres, instruments, moods, and production qualities—to audio tracks. Over the years, ad- vances in deep learning, an...

    work page

  2. [2]

    Methodology The methodology begins with feature extraction, followed by the grouping of related features to support the model’s under- standing 1. Finally, the EM- BANDED algorithm processes these grouped features to learn, analyze, and generate predictions, while simultaneously providing semantic explanations of its op- eration. An overview of the method...

    work page Pith review arXiv 2025

  3. [3]

    Experiments 3.1. Experimental Setup We conducted our experiments using three multimodal music datasets that include both audio and lyrics: two curated from Audioset and Music4All, as described in [7], resulting in single- label datasets; and the MTG Jamendo dataset [26], which re- quired preprocessing to extract lyrics (see Section 2.1). Mu- sic4All encom...

    work page

  4. [4]

    Conclusion & Future Work An important contribution of this work is demonstrating how feature grouping—whether User-Friendly, Domain-Expert, or otherwise—can be integrated seamlessly into a music tagging pipeline to deliver clear, high-level explanations. By leveraging our approach of EM- BANDED model we achieve to not only preserve good predictive perform...

    work page

  5. [5]

    Basic research Financing (Horizontal support of all Sciences)

    Acknowledgments The research project is implemented in the framework of H.F.R.I call “Basic research Financing (Horizontal support of all Sciences)” under the National Recovery and Resilience Plan “Greece 2.0” funded by the European Union –NextGenera- tionEU(H.F.R.I. Project Number: 15111 - Emotional Artificial Intelligence in Music Expression)

    work page

  6. [6]

    Foundation models for music: A survey,

    Y . Ma, A. Øland, A. Ragni, B. M. Del Sette, C. Saitis, C. Donahue, C. Lin, C. Plachouras, E. Benetos, E. Shatri et al., “Foundation models for music: A survey,” arXiv preprint arXiv:2408.14340 , 2024

    arXiv 2024

  7. [7]

    Mert: Acoustic music understanding model with large-scale self-supervised training,

    Y . Li, R. Yuan, G. Zhang, Y . Ma, X. Chen, H. Yin, C. Xiao, C. Lin, A. Ragni, E. Benetos et al., “Mert: Acoustic music understanding model with large-scale self-supervised training,” arXiv preprint arXiv:2306.00107, 2023

    arXiv 2023

  8. [8]

    Sample-level deep convolutional neural networks for music auto-tagging,

    J. Lee, J. Park, and J. Nam, “Sample-level deep convolutional neural networks for music auto-tagging,”IEEE Signal Processing Letters, 2019

    work page 2019

Show all 32 references
  1. [9]

    A foundation model for mu- sic informatics,

    M. Won, Y .-N. Hung, and D. Le, “A foundation model for mu- sic informatics,” in ICASSP 2024-2024 IEEE International Con- ference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 1226–1230

    2024

  2. [10]

    Towards a rigorous science of in- terpretable machine learning,

    F. Doshi-Velez and B. Kim, “Towards a rigorous science of in- terpretable machine learning,” arXiv preprint arXiv:1702.08608, 2017

    2017 arXiv

  3. [11]

    Toward interpretable music tagging with self-attention,

    M. Won, S. Chun, and X. Serra, “Toward interpretable music tagging with self-attention,” CoRR, vol. abs/1906.04972, 2019. [Online]. Available: http://arxiv.org/abs/1906.04972

    1906 arXiv

  4. [12]

    Musiclime: Explainable multimodal music understanding,

    T. Sotirou, V . Lyberatos, O. M. Mastromichalakis, and G. Sta- mou, “Musiclime: Explainable multimodal music understanding,” arXiv preprint arXiv:2409.10496, 2024

    2024 arXiv

  5. [13]

    Concept-based techniques for

    F. Foscarin, K. Hoedt, V . Praher, A. Flexer, and G. Widmer, “Concept-based techniques for” musicologist-friendly” explana- tions in a deep music classifier,”arXiv preprint arXiv:2208.12485, 2022

    2022 arXiv

  6. [14]

    Leveraging pre- trained autoencoders for interpretable prototype learning of music audio,

    P. Alonso-Jim ´enez, L. Pepino, R. Batlle-Roca, P. Zinemanas, D. Bogdanov, X. Serra, and M. Rocamora, “Leveraging pre- trained autoencoders for interpretable prototype learning of music audio,” arXiv preprint arXiv:2402.09318, 2024

    2024 arXiv

  7. [15]

    A model you can hear: Audio identification with playable prototypes,

    R. Loiseau, B. Bouvier, Y . Teytaut, E. Vincent, M. Aubry, and L. Landrieu, “A model you can hear: Audio identification with playable prototypes,” ISMIR, 2022

    2022

  8. [16]

    Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,

    C. Rudin, “Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,” Nature Machine Intelligence, vol. 1, no. 5, pp. 206–215, 2019

    2019

  9. [17]

    A data-driven approach to mid-level perceptual musical feature modeling,

    A. Aljanaki and M. Soleymani, “A data-driven approach to mid-level perceptual musical feature modeling,” in ISMIR, 2018. [Online]. Available: https://api.semanticscholar.org/CorpusID: 49190827

    2018

  10. [18]

    Towards Explainable Music Emotion Recognition: The Route via Mid-level Features

    S. Chowdhury, A. V . Portabella, V . Haunschmid, and G. Widmer, “Towards Explainable Music Emotion Recognition: The Route via Mid-level Features.” ISMIR, 2019. [Online]. Available: https://doi.org/10.5281/zenodo.3527788

    2019 doi

  11. [19]

    Per- ceptual musical features for interpretable audio tagging,

    V . Lyberatos, S. Kantarelis, E. Dervakos, and G. Stamou, “Per- ceptual musical features for interpretable audio tagging,” in 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW). IEEE, 2024, pp. 878–882

    2024

  12. [20]

    The role of structure in the mu- sical expression of emotions,

    A. Gabrielsson and E. Lindstr ¨om, “The role of structure in the mu- sical expression of emotions,” Handbook of music and emotion: Theory, research, applications, vol. 367400, pp. 367–44, 2010

    2010

  13. [21]

    A multidimensional study of perceptual-emotional qualities in music,

    L. Wedin, “A multidimensional study of perceptual-emotional qualities in music,” Scandinavian journal of psychology, 1972

    1972

  14. [22]

    Rhyme and style fea- tures for musical genre classification by song lyrics

    R. Mayer, R. Neumayer, and A. Rauber, “Rhyme and style fea- tures for musical genre classification by song lyrics.” in Ismir, vol. 14, no. 18, 2008, pp. 337–342

    2008

  15. [23]

    Lyrics-based analysis and classifica- tion of music,

    M. Fell and C. Sporleder, “Lyrics-based analysis and classifica- tion of music,” inProceedings of COLING 2014, the 25th interna- tional conference on computational linguistics: Technical papers, 2014, pp. 620–631

    2014

  16. [24]

    Es- sentia: an open-source library for sound and music analysis,

    D. Bogdanov, N. Wack, E. G ´omez, S. Gulati, P. Herrera, O. Mayor, G. Roma, J. Salamon, J. Zapata, and X. Serra, “Es- sentia: an open-source library for sound and music analysis,” in Proceedings of the 21st ACM international conference on Multi- media, 2013, pp. 855–858

    2013

  17. [25]

    Func- tional harmony ontology: Musical harmony analysis with descrip- tion logics,

    S. Kantarelis, E. Dervakos, N. Kotsani, and G. Stamou, “Func- tional harmony ontology: Musical harmony analysis with descrip- tion logics,” Journal of Web Semantics, p. 100754, 2022

    2022

  18. [26]

    Exploring an em-algorithm for banded regression in computational neuroscience,

    S. A. Fuglsang, K. H. Madsen, O. Puonti, H. R. Siebner, and J. Hjortkjær, “Exploring an em-algorithm for banded regression in computational neuroscience,” Imaging Neuroscience, vol. 2, pp. 1–19, 2024

    2024

  19. [27]

    Omnizart: A general toolbox for automatic music transcription,

    Y .-T. Wu, Y .-J. Luo, T.-P. Chen, I.-C. Wei, J.-Y . Hsu, Y .-C. Chuang, and L. Su, “Omnizart: A general toolbox for automatic music transcription,” Journal of Open Source Software, vol. 6, no. 68, p. 3391, 2021. [Online]. Available: https://doi.org/10.21105/joss.03391

    2021 doi

  20. [28]

    Hybrid transformers for music source separation,

    S. Rouard, F. Massa, and A. D ´efossez, “Hybrid transformers for music source separation,” in ICASSP 2023-2023 IEEE Interna- tional Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

    2023

  21. [29]

    Robust speech recognition via large-scale weak supervision,

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervision,” in International conference on machine learning . PMLR, 2023, pp. 28 492–28 518

    2023

  22. [30]

    Exploring an em-algorithm for banded regression in computational neuroscience,

    S. A. Fuglsang, K. H. Madsen, O. Puonti, H. R. Siebner, and J. Hjortkjær, “Exploring an em-algorithm for banded regression in computational neuroscience,” Imaging Neuroscience, vol. 2, pp. 1–19, 05 2024. [Online]. Available: https://doi.org/10.1162/ imag a 00155

    2024

  23. [31]

    The mtg-jamendo dataset for automatic music tagging

    D. Bogdanov, M. Won, P. Tovstogan, A. Porter, and X. Serra, “The mtg-jamendo dataset for automatic music tagging.” ICML, 2019

    2019

  24. [32]

    Automatic mu- sic genre classification based on modulation spectral analysis of spectral and cepstral features,

    C.-H. Lee, J.-L. Shih, K.-M. Yu, and H.-S. Lin, “Automatic mu- sic genre classification based on modulation spectral analysis of spectral and cepstral features,”IEEE Transactions on Multimedia, vol. 11, no. 4, pp. 670–682, 2009

    2009

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.