Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Leveraging Allophony in Self-Supervised Speech Models for Atypical Pronunciation Assessment

T0 review · 4 major / 6 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read The paper claims that scoring pronunciation with a 32-component Gaussian mixture over frozen self-supervised speech features captures allophonic variation and outperforms prior methods on four of five atypical-speech datasets.

desk verdict Plausible method, honest disclosure, but the four-of-five SOTA claim is likely a selection artifact. read the letter →

arxiv 2502.07029 v2 pith:ZIEA5EIC submitted 2025-02-10 cs.CL cs.AIcs.LGeess.AS

Kwanghee Choi , Eunjung Yeo , Kalvin Chang , Shinji Watanabe , David Mortensen This is my paper · ORCID
classification cs.CLcs.AIcs.LGeess.AS
keywords pronunciationassessmentallophonyGaussianmixturemodelself-supervisedspeechmodelsout-of-distributiondetectiondysarthricnon-nativegoodnessof
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 argues that atypical pronunciation assessment should model allophonic variation, the different acoustic realizations of the same phoneme, instead of collapsing each phoneme into a single cluster as phoneme-classifier Goodness-of-Pronunciation methods do. It introduces MixGoP, which fits a Gaussian mixture with 32 subclusters to the frozen features of a self-supervised speech model for each phoneme and scores a segment by its log-likelihood. On five datasets of dysarthric and non-native speech, MixGoP reports the highest Kendall-tau correlation with human pronunciation scores on four of them, with the largest gains on dysarthric speech. The paper also shows that self-supervised features carry more phonetic-environment information than MFCCs or Mel spectrograms, and that this extra information is associated with better downstream scores up to a saturation point.

What carries the argument

The load-bearing object is MixGoP, a Gaussian mixture model per phoneme: $$P_\$\theta$(s\mid p)=\sum_{c=1}^C \pi^c_p\,\mathcal{N}(\mathrm{Enc}(s);\mu^c_p,\Sigma^c_p)$$ with C=32 full-covariance Gaussians fit by k-means-initialized EM to center-pooled frozen self-supervised speech features. The score is the log-likelihood; because it omits the softmax and uses Mahalanobis distances, it functions as an out-of-distribution detector. The analysis introduces ANMI, the normalized mutual information between k-means cluster indices of a phoneme and the natural class of neighboring phonemes, as a way to quantify how much allophonic information a feature representation carries. The feature extractors are WavLM-Large and XLS-R-300M, with features taken layer-wise and center-pooled per phoneme segment.

What would settle it

Re-run MixGoP many times with different random 512-sample subsamples of the same training features and measure the spread of correlations with human scores; if the ranking of utterances changes substantially between subsamples, the likelihood estimates are dominated by estimation noise rather than allophonic structure.

Watch

Extended reading notes

Core claim

The central claim is that replacing the softmax-based phoneme posterior of conventional Goodness of Pronunciation with a per-phoneme Gaussian mixture likelihood computed on frozen self-supervised features improves atypical pronunciation assessment, because it lets each phoneme occupy multiple subclusters corresponding to allophonic variants and treats atypical speech as out-of-distribution rather than in-distribution. MixGoP is defined as log P(s|p), the log-likelihood of the segment under a 32-component Gaussian mixture trained on typical speech; after k-means initialization, EM fits the means, covariances, and mixing weights. The quadratic form inside each Gaussian is a Mahalanobis distance, directly connecting the score to out-of-distribution detection. With WavLM-Large or XLS-R features, MixGoP achieves the best Kendall-tau on UASpeech, TORGO, SSNCE, and speechocean762, while NN-GoP remains best on L2-ARCTIC; the paper attributes the dysarthric gains to those test sets being more strongly out-of-distribution relative to training speech.

Load-bearing premise

The load-bearing assumption is that a 32-component bell-curve mixture fit to at most 512 high-dimensional speech-feature samples per phoneme, with no covariance regularization, yields stable scores that rank atypical pronunciations the way human raters do.

Editorial extensions

If this is right

  • Phoneme-level pronunciation scoring no longer needs a trained phoneme classifier; a frozen self-supervised model plus a per-phoneme Gaussian mixture suffices.
  • Dysarthric assessment can expect larger gains from likelihood-based out-of-distribution scoring than non-native assessment, because dysarthric test speech is further from typical training speech.
  • Layer choice matters more for XLS-R than for WavLM, since XLS-R's later layers degrade sharply; practitioners should select layers by downstream validation rather than defaulting to the final layer.
  • The finding that S3M subclusters align with phonetic environments suggests that cluster size in discrete speech-unit pipelines can be tuned to capture allophonic granularity between phoneme and speaker information.
  • The sample-efficiency result, with performance plateauing near 512 samples per phoneme, indicates the method is usable in low-resource atypical-speech settings.

Reading between the lines

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

  • The ANMI saturation pattern suggests a cheap layer-selection heuristic: pick the earliest layer whose ANMI approaches the plateau, which the paper does not test directly.
  • Feeding the per-phoneme MixGoP log-likelihood as an extra feature into a small supervised ranker could combine out-of-distribution sensitivity with task-specific weights; the paper only uses the likelihood as the final score.
  • Because only English and Tamil are tested, the natural stress test is to measure ANMI and MixGoP on a language with rich allophony outside those families, such as a tone language.
  • The phoneme-attention extension is trained using test-set labels, so a held-out evaluation would be needed before treating those small gains as deployable.
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

4 major / 6 minor

Summary. The paper proposes MixGoP, a pronunciation-assessment score that replaces the phoneme-classifier posterior of conventional Goodness-of-Pronunciation with a per-phoneme Gaussian mixture likelihood over frozen self-supervised speech model (S3M) features. The motivation is that allophonic variation creates multimodal phoneme distributions, and that atypical speech should be treated as out-of-distribution relative to typical speech. The method is evaluated on five datasets (UASpeech, TORGO, SSNCE, speechocean762, L2-ARCTIC) using MFCC, Mel-spectrogram, TDNN-F, XLS-R, and WavLM features, and the paper reports state-of-the-art Kendall-tau correlations on four of the five datasets. The paper also introduces an analysis metric, ANMI, measuring mutual information between k-means cluster indices of phoneme features and the surrounding phonetic environment, and uses it to argue that S3M features capture allophonic variation better than traditional acoustic features.

Significance. If the empirical claims hold, MixGoP is a conceptually clean and practically useful contribution: it shows that a frozen S3M feature extractor plus a per-phoneme density model can compete with or beat phoneme-classifier-based GoP methods on atypical speech without any fine-tuning. The paper is transparent about its protocol, releases code, and compares a broad set of baselines and features, which are strengths. The allophony analysis is also a useful complement to existing layer-wise analyses of S3M features. However, the strength of the headline claim depends on the evaluation protocol: the reported numbers are maxima over test-set-selected layers, subsample sizes, and cluster counts, with no error bars or significance tests, and the margins over the strongest baselines are small.

major comments (4)
  1. [§3.6, Table 1, Appendix C.1] The state-of-the-art claim rests on selecting, for each method and dataset, the best of 25 S3M layers on the same test set used for evaluation. The Table 1 caption says 'the best performance across layers is displayed,' and Figure 5 shows per-layer results; this is equivalent to reporting a maximum over the test set. Since the winning margins are small (e.g., UASpeech WavLM 0.623 vs. kNN 0.606; TORGO XLS-R 0.713 vs. kNN 0.704; speechocean762 WavLM 0.539 vs. NN-GoP 0.535), the four-of-five SOTA claim may be a selection artifact. Please report results with a held-out layer-selection protocol (e.g., validation split or nested cross-validation) or provide per-layer significance testing and confidence intervals.
  2. [§5.2 and Appendix C.2] The subsample size of 512 and the number of GMM components C=32 are also chosen on test performance. Table 2 searches over {64, 128, 256, 512, full} and Table 3 searches over C in {4, 8, 16, 32, 64}, with the layer index fixed to the best-performing layer from §3.6, which itself was test-selected. The final Table 1 numbers are therefore maxima over a hyperparameter grid evaluated on the test set. Please report stability of the rankings under different hyperparameter choices, or use a nested evaluation protocol, and quantify uncertainty (e.g., bootstrap over utterances or speakers).
  3. [§2.3 and §4.2] The GMM density estimate is likely unstable in the regime used. Features are up to 1024-dimensional, each phoneme GMM has C=32 full-covariance components, and training uses at most 512 samples per phoneme; the number of free covariance parameters vastly exceeds the sample size. Section 4.2 further states that 'we observed few to no EM optimization steps due to high dimensionality,' so the fitted model is essentially a k-means-initialized density estimator. No covariance regularization or convergence checks are reported, so the log-likelihood values in Eq. (5) may be numerically unstable or dominated by initialization. Please report covariance regularization, use tied/diagonal covariance or a distance-based alternative, and provide a bootstrap or multiple-restart analysis to show that the likelihoods are stable.
  4. [§3.5 and Appendix A.1] For the dysarthric datasets, speaker-level intelligibility scores are assigned to every utterance of that speaker, and Kendall-tau is then computed over utterances. This creates non-independence: utterances from the same speaker are not independent samples, and the effective sample size is much smaller than the number of utterances (e.g., UASpeech has only 14 dysarthric speakers). Without a clustered permutation test or a speaker-level analysis, the reported correlations and the differences between methods may be overconfident. Please account for the repeated-measures structure in the evaluation.
minor comments (6)
  1. [Abstract / §1] In the sentence 'distinguish atypical (out-of-distribution; OOD) from atypical (in-distribution) pronunciations,' the second 'atypical' should be 'typical'; this typo appears in the abstract and in §1.
  2. [§4.1 and Figure 2] The text refers to 'the distribution of /2/ (/AH/ in ARPABET),' but the figure uses ARPABET labels; '/2/' appears to be a typo for '/AH/' or another ARPABET symbol and should be corrected.
  3. [§4.2] The claim that 'we observed few to no EM optimization steps due to high dimensionality' is used to justify treating k-means centroids as the final GMM centroids, but the mechanism by which high dimensionality prevents EM optimization is not explained; please provide a concrete reason or a diagnostic (e.g., number of EM iterations actually run, or a log-likelihood trace).
  4. [§4.2 and Figure 3] The ANMI metric depends on the k-means cluster count k=32, which is chosen without sensitivity analysis; a brief statement of how ANMI changes with k (e.g., for k=16 or k=64) would strengthen the allophony claim.
  5. [§5.1 and Figure 4] The claim of a positive correlation between ANMI and downstream performance until NMI≈0.72 is based on only two models (XLS-R and WavLM) and appears to be an informal visual trend; please state this as an observation rather than a fitted relationship, or provide a quantitative regression.
  6. [§3.4] The phoneme classifier is trained with 'a single learnable FC layer' for a maximum of 500 iterations, but no early stopping or hyperparameter search is described; please clarify whether the same settings were used for all baselines and datasets, and whether any tuning was performed.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: MixGoP is an empirical likelihood fit scored against external human annotations; a few self-citations and the k-means-based allophony metric are secondary, not load-bearing.

full rationale

MixGoP is a straightforward empirical construction: a per-phoneme Gaussian mixture is fit to frozen S3M features from typical speech, and its log-likelihood is correlated with external human dysfluency and intelligibility scores. No predicted quantity is defined in terms of a fitted constant, and the state-of-the-art claim is checked against independent baselines (GMM-GoP, NN-GoP, kNN, oSVM) on external datasets, so the main result would stand or fall independently of the authors' prior work. The paper cites the authors' own earlier results for auxiliary design choices: Section 2.3 appeals to Yeo et al. (2023a) for dropping the phoneme prior, and Section 3.6 cites Choi and Yeo (2022) to interpret why S3M distance structure suits Mahalanobis and Euclidean scoring. These citations are supportive but not the evidence for the central empirical comparison. Section 4.2's ANMI metric is the closest to self-referential: the paper states that MixGoP's final centroids are likely similar to the k-means initializations because few to no EM optimization steps are observed, and ANMI measures environment information in k-means clusters of the same features. However, this is a secondary interpretative analysis, not the derivation of the state-of-the-art result, and its correlation with downstream performance in Figure 4 is an external empirical relationship. The disclosed test-set selection of layer, subsample size, and cluster count is a methodological weakness that could inflate reported correlations, but it is not circular: the evaluated quantity is still an external label correlation. Accordingly, no specific circular reduction can be exhibited, and the appropriate finding is no significant circularity, with a score of 2 reflecting minor self-citation and the self-referential flavor of the allophony analysis rather than a circular derivation.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central results depend on test-set-selected hyperparameters (cluster count, subsample size, layer index) and a fixed ANMI cluster count; no invented entities are introduced. The allophony interpretation assumes the phonetic-environment proxy and the distance-structure properties of S3M features.

free parameters (4)
  • Number of GMM subclusters C = 32
    Selected by grid search over 4, 8, 16, 32, 64 on the test set (Appendix C.2); central to the allophony modeling claim.
  • Subsampling maximum per phoneme = 512
    Chosen among 64, 128, 256, 512, full based on test performance (Section 5.2).
  • S3M layer index = varies by dataset and model, best on test
    Table 1 reports the best-performing layer out of 25 for each S3M per dataset, selected on the test set (Appendix C.1).
  • ANMI cluster count k = 32
    Fixed to match MixGoP's C; determines the normalized mutual information values in the allophony analysis (Section 4.2).
assumptions (5)
  • standard math A sufficiently large Gaussian mixture can approximate any continuous density.
    Used to justify fixing C=32 across phonemes (Section 2.3, Appendix C.2).
  • domain assumption Frozen S3M features encode acoustic information as relative distances, so Mahalanobis and Euclidean distances are effective for OOD scoring.
    Invoked in Section 3.6 to explain why MixGoP and kNN work with S3M features; supported by self-cited prior work (Choi and Yeo, 2022).
  • domain assumption The phonetic environment of a phoneme, represented by natural classes of neighboring phonemes plus word boundaries, is a valid proxy for allophonic variation.
    Defines the ANMI metric in Section 4.2 and underpins the analysis that S3Ms capture allophony.
  • domain assumption Center pooling of per-phoneme S3M features retains allophonic information while reducing alignment-error effects.
    Adopted in Section 3.2; if pooling discards segment-internal variation, the GMM likelihoods may not reflect true allophones.
  • domain assumption Ground-truth intelligibility and mispronunciation labels in the five datasets are reliable enough for Kendall-tau evaluation.
    All headline numbers are correlations against these labels (Section 3.5).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Allophony in Self-Supervised Speech Models for Atypical Pronunciation Assessment." pith.science (2026). https://pith.science/paper/ZIEA5EIC

@misc{pith2026250207029,
  author       = {Pith},
  title        = {Pith review of: Leveraging Allophony in Self-Supervised Speech Models for Atypical Pronunciation Assessment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZIEA5EIC}},
  note         = {Machine review of arXiv:2502.07029}
}
read the original abstract

Allophony refers to the variation in the phonetic realization of a phoneme based on its phonetic environment. Modeling allophones is crucial for atypical pronunciation assessment, which involves distinguishing atypical from typical pronunciations. However, recent phoneme classifier-based approaches often simplify this by treating various realizations as a single phoneme, bypassing the complexity of modeling allophonic variation. Motivated by the acoustic modeling capabilities of frozen self-supervised speech model (S3M) features, we propose MixGoP, a novel approach that leverages Gaussian mixture models to model phoneme distributions with multiple subclusters. Our experiments show that MixGoP achieves state-of-the-art performance across four out of five datasets, including dysarthric and non-native speech. Our analysis further suggests that S3M features capture allophonic variation more effectively than MFCCs and Mel spectrograms, highlighting the benefits of integrating MixGoP with S3M features.

Figures

Figures reproduced from arXiv: 2502.07029 by the authors.

Figure 1
Figure 1. Summary of our method, MixGoP. We model the likelihood of each phoneme using a Gaussian mix￾ture, trained on typical speech (in-distribution), to cap￾ture allophonic variations. We then evaluate on atypical speech (out-of-distribution). The y-axis represents the log-likelihood of a phoneme, where lower values indi￾cate greater atypicality. Before the era of deep neural networks (DNNs), allophones were modeled for sp… view at source ↗
Figure 2
Figure 2. demonstrates one example, with the distribution of /2/ (/AH/ in ARPABET) and its en￾vironments of the healthy subset of TORGO. We observed multiple clusters for each phoneme, each with phonetically similar environments, which mo￾tivates the metric for quantifying allophony ability. 4.2 Quantifying Allophony Previous studies have found that S3M features model the phoneme distributions with multiple Z IY AH M HH Z IY … view at source ↗
Figure 4
Figure 4. Comparing phonetic environment information with the downstream task performance. 5 Analysis 5.1 Does capturing phonetic environment leads to better downstream performance? It is crucial to examine whether capturing phonetic environments (or allophones) actually improves downstream performance. To assess this, we com￾pare the amount of phonetic environment informa￾tion inside S3Ms and the actual downstream per￾forman… view at source ↗
Figures from the paper (2 more)
Figure 7
Figure 7. Figure 7: Learned attention scores and the performance improvement on UASpeech, TORGO, SSNCE, and spee￾chocean762. within spearman is made differentiable by soft sort￾ing (Blondel et al., 2020), following Blondel et al. (2020)’s implementation. We applied 5-fold cross-validation…
Figure 6
Figure 6. Figure 6: Downstream performances with varying num￾ber of clusters and S3M layer index. Relative perfor￾mance improvement is obtained by dividing the Kendall￾tau correlation with that of cluster size 1. Except for the later layers of XLS-R with extreme performance degra￾dation, …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Towards Temporally Explainable Dysarthric Speech Clarity Assessment

    eess.AS 2025-05 conditional novelty 6.0 of 10

    A therapist-annotated dysarthric speech dataset and a three-stage ASR framework show that Whisper-large localizes mispronunciations precisely, with substitution errors detected best and 70.1% of ASR error descriptions...

Reference graph

Works this paper leans on

70 extracted references · 67 canonical work pages · cited by 1 Pith paper

  1. [1]

    M Abdullah, M

    B. M Abdullah, M. M. Shaik, B. M \"o bius, and D. Klakow. 2023. An information-theoretic analysis of self-supervised discrete representations of speech. In Proc. Interspeech

    work page 2023

  2. [2]

    A. Babu, C. Wang, A. Tjandra, et al. 2022. XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale . In Proc. Interspeech

    work page 2022

  3. [3]

    Baevski, Y

    A. Baevski, Y. Zhou, A. Mohamed, and M. Auli. 2020. wav2vec 2.0: A framework for self-supervised learning of speech representations. In Proc. NeurIPS

    work page 2020

  4. [4]

    Bilmes et al

    J.A. Bilmes et al. 1998. A gentle tutorial of the em algorithm and its application to parameter estimation for gaussian mixture and hidden markov models. International Computer Science Institute

    work page 1998

  5. [5]

    Blondel, O

    M. Blondel, O. Teboul, Q. Berthet, and J. Djolonga. 2020. Fast differentiable sorting and ranking. In Proc. ICML

    work page 2020

  6. [6]

    X. Cao, Z. Fan, T Svendsen, and G. Salvi. 2024. A framework for phoneme-level pronunciation assessment using ctc. In Proc. Interspeech

    work page 2024

  7. [7]

    Chang, B

    X. Chang, B. Yan, K. Choi, et al. 2024. Exploring speech recognition, translation, and understanding with discrete speech units: A comparative study. In Proc. ICASSP

    work page 2024

  8. [8]

    Chao, T-H

    F-A. Chao, T-H. Lo, T-I. Wu, et al. 2022. 3m: An effective multi-view, multi-granularity, and multi-aspect modeling approach to english pronunciation assessment. In Proc. APSIPA ASC

    work page 2022

Show all 70 references
  1. [9]

    S. Chen, C. Wang, Z. Chen, et al. 2022. Wavlm: Large-scale self-supervised pre-training for full stack speech processing. IEEE JSTSP

    2022

  2. [10]

    Cheng, Z

    S. Cheng, Z. Liu, L. Li, et al. 2020. ASR-Free Pronunciation Assessment . In Proc. Interspeech

    2020

  3. [11]

    Choi, J-W

    K. Choi, J-W. Jung, and S. Watanabe. 2024 a . Understanding probe behaviors through variational bounds of mutual information. In Proc. ICASSP

    2024

  4. [12]

    K. Choi, A. Pasad, T. Nakamura, et al. 2024 b . Self-supervised speech representations are more phonetic than semantic. In Proc. Interspeech

    2024

  5. [13]

    Choi and E-J

    K. Choi and E-J. Yeo. 2022. Opening the black box of wav2vec feature encoder. arXiv preprint arXiv:2210.15386

    2022 arXiv

  6. [14]

    Collins, I

    B. Collins, I. M Mees, and P. Carley. 2019. Phoneme, allophone and syllable. In Practical English Phonetics and Phonology. Routledge

    2019

  7. [15]

    H. Do, Y. Kim, and G-G. Lee. 2023. Hierarchical pronunciation assessment with multi-aspect attention. In Proc. ICASSP

    2023

  8. [16]

    P. Enderby. 1980. Frenchay dysarthria assessment. British Journal of Disorders of Communication

    1980

  9. [17]

    T-H. Feng, A. Dong, C-F Yeh, et al. 2023. SUPERB @ SLT 2022: Challenge on generalization and efficiency of self-supervised speech representation learning . In Proc. SLT

    2023

  10. [18]

    Y. Gong, Z. Chen, I-H. Chu, P. Chang, and J. Glass. 2022. Transformer-based multi-aspect multi-granularity non-native english speaker pronunciation assessment. In Proc. ICASSP

    2022

  11. [19]

    Hendrycks and K

    D. Hendrycks and K. Gimpel. 2017. A baseline for detecting misclassified and out-of-distribution examples in neural networks. In Proc. ICLR

    2017

  12. [20]

    Hernandez, E.J

    A. Hernandez, E.J. Yeo, S. Kim, and M. Chung. 2020. Dysarthria detection and severity assessment using rhythm-based metrics. In Proc. Interspeech

    2020

  13. [21]

    W-N. Hsu, B. Bolte, Yao-Hung H. Tsai, et al. 2021. HuBERT : Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM TASLP

    2021

  14. [22]

    W. Hu, Y. Qian, and F. K. Soong. 2015 a . An improved dnn-based approach to mispronunciation detection and diagnosis of l2 learners' speech. In SLaTE

    2015

  15. [23]

    W. Hu, Y. Qian, F. K Soong, and Y. Wang. 2015 b . Improved mispronunciation detection with deep neural network trained acoustic models and transfer learning based logistic regression classifiers. Speech Communication

    2015

  16. [24]

    Jokisch, A

    O. Jokisch, A. Wagner, R. Sabo, et al. 2009. Multilingual speech data collection for the assessment of pronunciation and prosody in a language learning system. In Proc. of SPECOM

    2009

  17. [25]

    H. Kim, M. Hasegawa-Johnson, A. Perlman, et al. 2008. Dysarthric speech database for universal access research. In Proc. Interspeech

    2008

  18. [26]

    H. Kim, K. Martin, M. Hasegawa-Johnson, and A. Perlman. 2010. Frequency of consonant articulation errors in dysarthric speech. Clinical linguistics & phonetics

    2010

  19. [27]

    D. P. Kingma and J. Ba. 2015. Adam: A method for stochastic optimization. In Proc. ICLR

    2015

  20. [28]

    Kominek and A

    J. Kominek and A. W. Black. 2004. The cmu arctic speech databases. In Proc. Speech Synthesis Workshop

    2004

  21. [29]

    Korzekwa, J

    D. Korzekwa, J. Lorenzo-Trueba, S. Zaporowski, S. Calamaro, T. Drugman, and B. Kostek. 2021. Mispronunciation detection in non-native (l2) english with uncertainty modeling. In Proc. ICASSP

    2021

  22. [30]

    Ladefoged

    P. Ladefoged. 1965. The nature of general phonetic theories. Monograph Series on Languages and Linguistics, pages 27--42

    1965

  23. [31]

    J. Lee, K. Lee, H. Lee, and J. Shin. 2018. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. In Proc. NeurIPS

    2018

  24. [32]

    K-F. Lee, S. Hayamizu, H-W. Hon, et al. 1990. Allophone clustering for continuous speech recognition. In Proc. ICASSP

    1990

  25. [33]

    K. Li, X. Qian, and H. Meng. 2016. Mispronunciation detection and diagnosis in l2 english speech using multidistribution deep neural networks. IEEE/ACM TASLP

    2016

  26. [34]

    O. D. Liu, H. Tang, and S. Goldwater. 2023. Self-supervised predictive coding models encode speaker and phonetic information in orthogonal subspaces. In Proc. Interspeech

    2023

  27. [35]

    MacKenzie and D

    L. MacKenzie and D. Turton. 2020. Assessing the accuracy of existing forced alignment software on varieties of british english. Linguistics Vanguard, 6(s1):20180061

    2020

  28. [36]

    Martin, J

    K. Martin, J. Gauthier, C. Breiss, and R. Levy. 2023. Probing Self-supervised Speech Models for Phonetic and Phonemic Information: A Case Study in Aspiration . In Proc. Interspeech

    2023

  29. [37]

    McAuliffe, M

    M. McAuliffe, M. Socolof, S. Mihuc, M. Wagner, and M. Sonderegger. 2017. Montreal forced aligner: Trainable text-speech alignment using kaldi. In Proc. Interspeech

    2017

  30. [38]

    McFee, C

    B. McFee, C. Raffel, D. Liang, et al. 2015. librosa: Audio and music signal analysis in python. In Proc. Python in Science Conference

    2015

  31. [39]

    McInnes, J

    L. McInnes, J. Healy, N. Saul, and L. Gro berger. 2018. Umap: Uniform manifold approximation and projection. The Journal of Open Source Software

    2018

  32. [40]

    B. K. Ng and P. S. Chiew. 2023. L1 influence on stop consonant production: A case study of malaysian mandarin-english bilinguals. Asian Englishes

    2023

  33. [41]

    Nguyen, H

    T. Nguyen, H. D. Nguyen, F. Chamroukhi, and G. J. McLachlan. 2020. Approximation by finite mixtures of continuous density functions that vanish at infinity. Cogent Mathematics & Statistics

    2020

  34. [42]

    Pasad, C-M

    A. Pasad, C-M. Chien, S. Settle, et al. 2024. What Do Self-Supervised Speech Models Know About Words? TACL

    2024

  35. [43]

    Pasad, J-C Chou, and K

    A. Pasad, J-C Chou, and K. Livescu. 2021. Layer-wise analysis of a self-supervised speech representation model. In Proc. ASRU, pages 914--921

    2021

  36. [44]

    Pasad, B

    A. Pasad, B. Shi, and K. Livescu. 2023. Comparative layer-wise analysis of self-supervised speech models. In Proc. ICASSP

    2023

  37. [45]

    Pedregosa, G

    F. Pedregosa, G. Varoquaux, A. Gramfort, et al. 2011. Scikit-learn: Machine learning in P ython. Journal of Machine Learning Research

    2011

  38. [46]

    Povey, G

    D. Povey, G. Cheng, Y. Wang, et al. 2018. Semi-orthogonal low-rank matrix factorization for deep neural networks. In Proc. Interspeech, pages 3743--3747

    2018

  39. [47]

    Rudzicz, A

    F. Rudzicz, A. K. Namasivayam, and T. Wolff. 2012. The torgo database of acoustic and articulatory speech from speakers with dysarthria. Language Resources and Evaluation

    2012

  40. [48]

    Sagayama

    S. Sagayama. 1989. Phoneme environment clustering for speech recognition. In Proc. ICASSP

    1989

  41. [49]

    Sch \"o lkopf, J

    B. Sch \"o lkopf, J. C Platt, J. Shawe-Taylor, et al. 2001. Estimating the support of a high-dimensional distribution. Neural computation

    2001

  42. [50]

    Shahin and B

    M. Shahin and B. Ahmed. 2019. Anomaly detection based pronunciation verification approach using speech attribute features. Speech Communication

    2019

  43. [51]

    Shahin, J

    M. Shahin, J. Epps, and B. Ahmed. 2024. Phonological level wav2vec2-based mispronunciation detection and diagnosis method. In Proc. Interspeech

    2024

  44. [52]

    J. Shi, N. Huo, and Q. Jin. 2020. Context-aware goodness of pronunciation for computer-assisted pronunciation training. In Proc. Interspeech

    2020

  45. [53]

    Sicherman and Y

    A. Sicherman and Y. Adi. 2023. Analysing discrete self supervised speech representation for spoken language modeling. In Proc. ICASSP

    2023

  46. [54]

    Sudhakara, M

    S. Sudhakara, M. K. Ramanathi, C. Yarra, and P. K. Ghosh. 2019. An improved goodness of pronunciation (gop) measure for pronunciation evaluation with dnn-hmm system considering hmm transition probabilities. In Proc. Interspeech

    2019

  47. [55]

    Y. Sun, Y. Ming, X. Zhu, and Y. Li. 2022. Out-of-distribution detection with deep nearest neighbors. In Proc. ICML

    2022

  48. [56]

    M. C. TA, T. Nagarajan, and P. Vijayalakshmi. 2016. Dysarthric speech corpus in tamil for rehabilitation research. In Proc. TENCON

    2016

  49. [57]

    William F Twaddell. 1952. Phonemes and allophones in speech analysis. Journal of the Acoustical Society of America

    1952

  50. [58]

    Vidal, L

    J. Vidal, L. Ferrer, and L. Brambilla. 2019. Epadb: A database for development of pronunciation assessment systems. In Proc. Interspeech

    2019

  51. [59]

    Z. Wang, Q. Gu, Y. Ning, and H. Liu. 2015. High dimensional em algorithm: Statistical optimization and asymptotic normality. In Proc. NeurIPS

    2015

  52. [60]

    Wells, H

    D. Wells, H. Tang, and K. Richmond. 2022. Phonetic Analysis of Self-supervised Representations of English Speech . In Proc. Interspeech

    2022

  53. [61]

    S. M. Witt and S. J. Young. 2000. Phone-level pronunciation scoring and assessment for interactive language learning. Speech Communication, 30:95--108

    2000

  54. [62]

    X. Xu, Y. Kang, S. Cao, et al. 2021. Explore wav2vec 2.0 for Mispronunciation Detection . In Proc. Interspeech

    2021

  55. [63]

    Yang, P-H

    S-W. Yang, P-H. Chi, Y-S. Chuang, et al. 2021. SUPERB: Speech Processing Universal PERformance Benchmark . In Proc. Interspeech, pages 1194--1198

    2021

  56. [64]

    E-J. Yeo, K. Choi, S. Kim, and M. Chung. 2023 a . Speech Intelligibility Assessment of Dysarthric Speech by using Goodness of Pronunciation with Uncertainty Quantification . In Proc. Interspeech

    2023

  57. [65]

    E-J. Yeo, H. Ryu, J. Lee, et al. 2023 b . Comparison of l2 korean pronunciation error patterns from five l1 backgrounds by using automatic phonetic transcription. In Proc. ICPhS

    2023

  58. [66]

    S. J. Young, J. J. Odell, and P. C. Woodland. 1994. Tree-based state tying for high accuracy modelling. In Proc. HLT

    1994

  59. [67]

    Zhang, Z

    J. Zhang, Z. Zhang, Y. Wang, et al. 2021. speechocean762: An open-source non-native english speech corpus for pronunciation assessment. In Proc. Interspeech

    2021

  60. [68]

    G. Zhao, S. Sonsaat, A. Silpachai, et al. 2018. L2-arctic: A non-native english speech corpus. In Proc. Interspeech

    2018

  61. [69]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  62. [70]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.