Pith. sign in

REVIEW 5 major objections 5 minor 65 references

Personalized Fine-Tuning with Controllable Synthetic Speech from LLM-Generated Transcripts for Dysarthric Speech Recognition

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a frozen Whisper model personalized with x-vector speaker embeddings and AdaLoRA adapters, trained with synthetic Parler-TTS speech from LLM-generated transcripts, lowers dysarthric word error rate from 12.31 to…

desk verdict A useful, honest challenge submission whose synthetic-data gain is only shown on dev; the authors disclose this, but the abstract and conclusions overreach, so the headline claim needs test1 scores before it should be trusted. read the letter →

arxiv 2505.12991 v1 pith:R6QIF4ZR submitted 2025-05-19 cs.SD eess.AS

Dominik Wagner , Ilja Baumann , Natalie Engert , Seanie Lee , Elmar Nöth , Korbinian Riedhammer , Tobias Bocklet This is my paper · ORCID
classification cs.SDeess.AS
keywords dysarthricspeechrecognitionpersonalizationx-vectorsAdaLoRAparameter-efficientfine-tuningsyntheticParler-TTSwav2vec2.0
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

This paper is trying to establish that dysarthric speech recognition can be substantially improved by combining three lightweight ingredients: speaker identity vectors (x-vectors) that condition a frozen speech model, parameter-efficient adapters (AdaLoRA), and synthetic training speech generated by a controllable text-to-speech model. The synthetic audio is produced by fine-tuning Parler-TTS on dysarthric recordings, with target transcripts written by a large language model so that the new data matches the topic and vocabulary of the real corpus. On a public dysarthric corpus, the authors report that this combination lowers word error rate from 12.31 on the unadapted Whisper baseline to 7.47 when synthetic speech equal in size to the real training set is added. They also report that synthetic data helps most for speakers with ALS, and that the best synthetic-data models were not evaluated on the held-out test set.

What carries the argument

The mechanism that carries the argument is speaker conditioning injected into the decoder. For each utterance, an x-vector embedding is averaged over frames and passed through a two-layer mapping network into the decoder's hidden dimension; the projected vector is concatenated with the encoder output before decoding. Because only these mapping networks and AdaLoRA adapters are trained, personalization is parameter-efficient. The synthetic-data half of the argument is carried by Parler-TTS, an autoregressive text-to-speech model that conditions on natural-language descriptions of voice attributes; it is fine-tuned on dysarthric speech, and its target transcripts are generated by an LLM prompted with ten random utterances from the corpus so the synthetic text stays in-domain. Synthetic utterances that a Whisper model transcribes with WER above 35 are discarded.

What would settle it

Score the 100%-synthetic-data model and the personalized AdaLoRA baseline on the held-out test set; if the 100%-synthetic model does not beat the baseline there, the claim that synthetic speech helps is unsupported, since the dev-set ranking would not transfer.

Watch

Extended reading notes

Core claim

The central claim is that data scarcity, not model capacity, is the main obstacle to dysarthric ASR, and that a modular pipeline can compensate. Starting from Whisper, the authors add x-vector speaker embeddings that are projected into the decoder and concatenated with encoder outputs; they train only AdaLoRA adapters and the mapping networks. With x-vector personalization, AdaLoRA reaches 8.05 WER on the dev set versus 11.73 without personalization. Adding latent representations from the last layer of a wav2vec 2.0 model lowers WER to 7.68, and adding 100% synthetic Parler-TTS speech produced from LLM-generated transcripts lowers it further to 7.47. The authors report that synthetic speech from a fine-tuned TTS model outperforms zero-shot voice-cloning synthetic data, and that the benefit is not uniform across etiologies.

Load-bearing premise

The central claim assumes that word error rates measured on the development set predict performance on the unseen test set; the authors state that their best synthetic-data models were never scored on the test set, so the reported improvement could be specific to the dev distribution.

Editorial extensions

If this is right

  • A frozen large speech model can be adapted to a dysarthric speaker with only small adapters and a speaker vector, avoiding full retraining and its data requirements.
  • A controllable TTS model fine-tuned on the target disorder can supply useful training data when real dysarthric recordings are scarce.
  • LLM-generated transcripts that imitate the topic and phrasing of the real corpus add diversity without leaving the domain, which matters when the available transcript inventory is small.
  • Synthetic-data gains are condition-dependent, so deployment systems should weigh augmentation by etiology rather than expecting a uniform benefit.
  • Additional self-supervised audio representations (wav2vec 2.0) help in the personalized setting but did not improve further when combined with synthetic data in the tested configurations.

Reading between the lines

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

  • One testable implication is that the text descriptions used to prompt Parler-TTS could be varied systematically to isolate which voice attributes (pitch, rate, breathiness) drive the ASR gains, turning the generator into a diagnostic for dysarthric speech.
  • The same pipeline may transfer to other atypical speech conditions, such as stuttering or accented speech, since neither the adapter personalization nor the LLM transcript generation is specific to dysarthria.
  • Given the large dev-to-test gap on the few configurations that were scored on the test set, a checkpoint-selection rule based on a speaker-holdout rather than the dev set may generalize better; this is an inference from the paper's reported numbers, not a claim it makes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper presents a system for dysarthric speech recognition submitted to the Speech Accessibility Project challenge. The authors build on a Whisper-based encoder-decoder architecture with x-vector-based personalization and parameter-efficient fine-tuning, comparing full fine-tuning, LoRA, and AdaLoRA. They extend the model by adding latent wav2vec 2.0 representations and by generating synthetic dysarthric speech through a fine-tuned Parler-TTS model conditioned on LLM-generated transcripts. Experiments on the SAP dataset report dev-set WER improvements from personalization, AdaLoRA, wav2vec2 features, and synthetic data, with the best dev WER of 7.47 compared to 8.05 for the personalized AdaLoRA baseline. Test1 scores are reported only for a subset of configurations and are substantially higher than the corresponding dev scores. The paper includes a limitations section that explicitly acknowledges that the best synthetic-data models were not scored on the unseen test1 set.

Significance. If the dev-set improvements transfer to the held-out test set, the combination of x-vector personalization, AdaLoRA, and wav2vec2 features would be a practically useful and lightweight recipe for dysarthric ASR, where training data are scarce. The paper provides a detailed experimental configuration that supports reproducibility, a systematic comparison of adaptation methods, and an honest limitations section. The main contribution of controllable synthetic data generation via LLM-based transcripts and Parler-TTS is interesting but currently rests almost entirely on dev-set evidence, so the significance of that specific component is not yet established.

major comments (5)
  1. [Table 3 and Section 4.5] The abstract and Section 5 claim that synthetic data yields up to ~7% relative WER improvement over personalized fine-tuning alone (comparing #10 at 8.05 and #2.3 at 7.47). This claim is based entirely on dev-set results, as Section 4.5 states that the best synthetic-data models were not scored on test1. The test1 scores that are available for synthetic-data configurations show dev-to-test degradations at least as large as those of the non-synthetic baseline #10 (dev 8.05 vs test1 10.65): #2.4 (dev 9.28 vs test1 11.68), #2.5 (dev 8.49 vs test1 11.68), and #2.14 (dev 8.35 vs test1 12.61). Without test1 scores for #2.1-2.3, the generalization of the synthetic-data benefit is unverified; please either provide test1 scores for these key configurations or temper the corresponding claims in the abstract and conclusions.
  2. [Section 4] Hyperparameters and checkpoints are selected on the dev set: the text states results are reported under the best hyperparameter configuration on the dev set, and the best model is chosen based on the lowest WER on a randomly sampled ~10% subset of the dev set. This selection procedure makes the reported dev WERs optimistically biased. No error bars, confidence intervals, or significance tests are provided. The key differences, such as #10 (8.05) versus #1.3 (7.68) and #2.3 (7.47), are small in absolute terms (0.21-0.58 WER points) and could fall within run-to-run variability. Please add multi-run statistics or at least a paired significance test for the main comparisons.
  3. [Table 2 and Section 4.3] The claimed ~5% relative improvement from wav2vec2 audio representations (#10 at 8.05 vs #1.3 at 7.68) is also dev-only. The reported test1 scores for the related configurations #1.4 (10.78) and #1.5 (10.67) are essentially identical to the test1 score of #10 (10.65), providing no evidence that this component helps on unseen data. Please report test1 scores for the best wav2vec2 configuration or scale back the claim.
  4. [Abstract and Table 1] The abstract states that AdaLoRA outperforms full fine-tuning and standard low-rank adaptation with relative WER reductions of ~23% and ~22%, respectively. From Table 1, the personalized AdaLoRA result (#10, WER 8.05) reduces WER by 23.5% relative to personalized LoRA (#8, WER 10.52) but by 30.2% relative to personalized FFT (#12, WER 11.54). Please correct the stated percentages or clarify the exact comparison used.
  5. [Section 4, synthetic data filtering] The synthetic data pipeline filters out Parler-TTS utterances whose Whisper-medium transcription has a WER of 35 or higher. This biases the retained synthetic training data toward utterances that a mid-size general ASR model already decodes well, which may not represent the most challenging dysarthric speech. The paper does not report how many utterances were discarded or whether the WER distribution of the retained synthetic data matches the real SAP training data. Please quantify this filtering effect and discuss its potential influence on the synthetic-data results.
minor comments (5)
  1. [Section 4.3] The sentence 'The lower half of the table, where personalization is is not used' contains a duplicated 'is'; it should read '...where personalization is not used...'.
  2. [Table 3] For experiments #2.10-2.12, the table marks 'Audio Repr.' but does not specify which wav2vec2 layer was used; the text says these follow #1.3, so the table should state 'base (layer 12)' for clarity.
  3. [Section 4.4 and Table 3] The TTS model name 'OpenVoice' is rendered as 'OpenV oice' in the text and table; please fix the typographical spacing.
  4. [Section 2] The dataset description reports an average utterance duration of 7.9 +/- 9.0 seconds; please clarify whether the +/- term is a standard deviation, since the presented format could be mistaken for a confidence interval.
  5. [Section 4.1] The test1 scores are described as 'non-public'; consider adding a sentence about the access conditions for these scores, as this affects the reproducibility of the test1 numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical WER/SemScore comparisons, not derivations that reduce to their inputs.

full rationale

This is an empirical systems paper rather than a derivation. Every central claim (x-vector personalization, AdaLoRA over LoRA/FFT, wav2vec 2.0 feature integration, and synthetic-data gains) is supported by WER/SemScore comparisons on the SAP dev set and, for some runs, the unseen test1 set. The authors adopt their own prior architecture [33] as the starting point, but the comparison between experiments #4 and #10 provides independent empirical evidence for the personalization benefit within this paper, so the self-citation is not load-bearing. Likewise, the synthetic-data pipeline is a data-augmentation method: Parler-TTS is fine-tuned on the training data and used to generate additional training examples, but the reported improvement is measured on dev/test, not derived from the generation procedure. Filtering synthetic utterances with Whisper-medium may bias the synthetic data toward recognizable examples, but that is a validity and generalization concern, not a circularity in which a predicted quantity is equivalent to an input by construction. The absence of test1 scores for the best synthetic-data configurations is an evidence-limitation issue, explicitly disclosed in Section 4.5, and does not constitute circular reasoning. No equation in the paper defines a target quantity in terms of itself, and no fitted parameter is renamed as a prediction. The paper is therefore self-contained as an empirical study, and the circularity score is 0.

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

The empirical claims rest on standard ML assumptions about pretrained model transfer, speaker embeddings, and synthetic-data domain match, plus the assumption that dev performance predicts test performance, which the paper itself questions. No new physical or mathematical entities are introduced, so invented_entities is empty.

free parameters (4)
  • Hyperparameter configuration per training method = Not reported (searched ranges in Section 4)
    The best configuration for LoRA, AdaLoRA, and FFT is selected on dev WER, so reported gains are conditional on dev-set selection.
  • Synthetic data ratio = 100%
    Best of 10%, 50%, and 100% on dev for Parler-TTS without SpecAugment (Table 3).
  • wav2vec 2.0 feature layer = 12
    Best of layers 2, 6, and 12 on dev (Table 2).
  • Synthetic utterance filter threshold = WER < 35
    Ad hoc threshold; synthetic utterances transcribed by Whisper-medium with WER at or above 35 are discarded (Section 4).
assumptions (4)
  • domain assumption Whisper-large-v3, frozen except adapters, transfers to dysarthric speech after adaptation.
    The system relies on Whisper's pretrained representations as the backbone (Section 3.1).
  • domain assumption x-vectors extracted from a VoxCeleb-trained model capture speaker-specific dysarthric characteristics useful for ASR.
    Used as personalization input (Section 3.1, Equations 3-5).
  • domain assumption Parler-TTS, fine-tuned on SAP training data, can generate synthetic dysarthric speech whose distribution matches enough of the real data to improve ASR training.
    Core to the synthetic data augmentation claim (Section 3.2).
  • domain assumption SAP dev set WER is a valid proxy for unseen test performance.
    Model selection and all headline results use dev (Section 4); the paper itself questions this in Section 4.5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Personalized Fine-Tuning with Controllable Synthetic Speech from LLM-Generated Transcripts for Dysarthric Speech Recognition." pith.science (2026). https://pith.science/paper/R6QIF4ZR

@misc{pith2026250512991,
  author       = {Pith},
  title        = {Pith review of: Personalized Fine-Tuning with Controllable Synthetic Speech from LLM-Generated Transcripts for Dysarthric Speech Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R6QIF4ZR}},
  note         = {Machine review of arXiv:2505.12991}
}
read the original abstract

In this work, we present our submission to the Speech Accessibility Project challenge for dysarthric speech recognition. We integrate parameter-efficient fine-tuning with latent audio representations to improve an encoder-decoder ASR system. Synthetic training data is generated by fine-tuning Parler-TTS to mimic dysarthric speech, using LLM-generated prompts for corpus-consistent target transcripts. Personalization with x-vectors consistently reduces word error rates (WERs) over non-personalized fine-tuning. AdaLoRA adapters outperform full fine-tuning and standard low-rank adaptation, achieving relative WER reductions of ~23% and ~22%, respectively. Further improvements (~5% WER reduction) come from incorporating wav2vec 2.0-based audio representations. Training with synthetic dysarthric speech yields up to ~7% relative WER improvement over personalized fine-tuning alone.

Figures

Figures reproduced from arXiv: 2505.12991 by the authors.

Figure 1
Figure 1. provides an overview of the components used in this work. The speech recognition system 1 is based on the ap￾proach introduced in [33]. It employs Whisper [34] as the ASR backbone and a personalization component based on a small neural network 3 , which projects audio representations ob￾tained from a pre-trained and frozen embedding model (e.g. x￾vectors [22]) 2 into the latent space of the decoder. The model is tra… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 54 canonical work pages

  1. [1]

    This impairment leads to difficulties with articulation, phonation, respiration, resonance, and prosody [1]

    Introduction Dysarthria is a motor speech disorder characterized by impaired control of the muscles involved in speech production, which can result from damage to the central or peripheral nervous system. This impairment leads to difficulties with articulation, phonation, respiration, resonance, and prosody [1]. Dysarthria manifests differently depending ...

    work page

  2. [2]

    The2024-11-30version of the data contains speech recordings from a total of 430 participants, with 374 assigned to the training set and 56 to the development (dev) set

    Data We employ the Speech Accessibility Project [37] (SAP) dataset for all experiments, with models submitted to the associated challenge.1 Specifically, we use the SAP2024-11-30data package, a dataset curated for the development of dysarthric ASR systems, which includes English dysarthric speech from individuals with PD, ALS, cerebral palsy, down syndrom...

    work page

  3. [3]

    The speech recognition system 1 is based on the ap- proach introduced in [33]

    Method Figure 1 provides an overview of the components used in this work. The speech recognition system 1 is based on the ap- proach introduced in [33]. It employs Whisper [34] as the ASR backbone and a personalization component based on a small neural network 3 , which projects audio representations ob- tained from a pre-trained and frozen embedding mode...

    work page

  4. [4]

    SemScore is a weighted sum of BERTscore [49], phonetic distance, and natural language infer- ence probability [50]

    Experiments and results Our experiments were evaluated using the SAP challenge eval- uation scheme, which is based on two metrics: WER and se- mantic score (SemScore). SemScore is a weighted sum of BERTscore [49], phonetic distance, and natural language infer- ence probability [50]. For each experiment, we explored mul- tiple hyperparameter configurations...

    work page

  5. [5]

    We fine-tuned Parler-TTS to generate control- lable synthetic dysarthric speech using LLM-generated tran- scripts and incorporated personalization via x-vectors

    Conclusions This work explored improving dysarthric speech recognition through personalized fine-tuning with speaker vectors, synthetic speech augmentation, and the integration of additional audio representations. We fine-tuned Parler-TTS to generate control- lable synthetic dysarthric speech using LLM-generated tran- scripts and incorporated personalizat...

    work page

  6. [6]

    Whistle-blowing ASRs: Evaluating the need for more inclusive speech recognition systems,

    M. Moore, H. Venkateswara, and S. Panchanathan, “Whistle-blowing ASRs: Evaluating the need for more inclusive speech recognition systems,” inInterspeech, 2018, pp. 466–470

    work page 2018

  7. [7]

    Differential diagnostic patterns of dysarthria,

    F. L. Darley, A. E. Aronson, and J. R. Brown, “Differential diagnostic patterns of dysarthria,”Journal of Speech and Hearing Research, vol. 12, no. 2, pp. 246–269, 1969

    work page 1969

  8. [8]

    Parkinson’s disease,

    A. E. Lang and A. M. Lozano, “Parkinson’s disease,”New England Journal of Medicine, vol. 339, no. 15, pp. 1044–1053, 1998

    work page 1998

Show all 65 references
  1. [9]

    Ageing and parkinson’s disease: substantia nigra regional selectivity,

    J. M. Fearnley and A. J. Lees, “Ageing and parkinson’s disease: substantia nigra regional selectivity,”Brain, vol. 114, no. 5, pp. 2283–2301, 10 1991

    1991

  2. [10]

    Speech and swallowing symptoms associ- ated with parkinson’s disease and multiple sclerosis: A survey,

    L. Hartelius and P. Svensson, “Speech and swallowing symptoms associ- ated with parkinson’s disease and multiple sclerosis: A survey,”Folia Pho- niatrica et Logopaedica, vol. 46, no. 1, 12 2009

    2009

  3. [11]

    Individual manifestations of parkinson’s disease after ten or more years of levodopa,

    H. L. Klawans, “Individual manifestations of parkinson’s disease after ten or more years of levodopa,”Movement Disorders, vol. 1, no. 3, pp. 187– 192, 1986

    1986

  4. [12]

    Investiga- tion of data augmentation techniques for disordered speech recognition,

    M. Geng, X. Xie, S. Liu, J. Yu, S. Hu, X. Liu, and H. Meng, “Investiga- tion of data augmentation techniques for disordered speech recognition,” in Interspeech, 2020, pp. 696–700

    2020

  5. [14]

    Training data augmen- tation for dysarthric automatic speech recognition by text-to-dysarthric- speech synthesis,

    W.-Z. Leung, M. Cross, A. Ragni, and S. Goetze, “Training data augmen- tation for dysarthric automatic speech recognition by text-to-dysarthric- speech synthesis,” inInterspeech, 2024

    2024

  6. [15]

    Combining in-domain and out-of-domain speech data for automatic recognition of disordered speech,

    H. Christensen, M. B. Aniol, P. Bell, P. D. Green, T. Hain, S. King, and P. Swietojanski, “Combining in-domain and out-of-domain speech data for automatic recognition of disordered speech,” inInterspeech, 2013

    2013

  7. [16]

    On the impact of dysarthric speech on contem- porary ASR cloud platforms,

    L. De Russis and F. Corno, “On the impact of dysarthric speech on contem- porary ASR cloud platforms,”Journal of Reliable Intelligent Environments, vol. 5, no. 3, pp. 163–172, Sep. 2019

    2019

  8. [17]

    Data augmentation using healthy speech for dysarthric speech recognition,

    B. Vachhani, C. Bhat, and S. K. Kopparapu, “Data augmentation using healthy speech for dysarthric speech recognition,” inInterspeech, 2018

    2018

  9. [18]

    Adver- sarial data augmentation using vae-gan for disordered speech recognition,

    Z. Jin, X. Xie, M. Geng, T. Wang, S. Hu, J. Deng, G. Li, and X. Liu, “Adver- sarial data augmentation using vae-gan for disordered speech recognition,” inICASSP, 2023, pp. 1–5

    2023

  10. [19]

    Improved ASR performance for dysarthric speech using two-stage data augmentation,

    C. Bhat, A. Panda, and H. Strik, “Improved ASR performance for dysarthric speech using two-stage data augmentation,” inInterspeech, 2022, pp. 46–50

    2022

  11. [20]

    Phonetic analysis of dysarthric speech tempo and applications to robust personalised dysarthric speech recognition,

    F. Xiong, J. Barker, and H. Christensen, “Phonetic analysis of dysarthric speech tempo and applications to robust personalised dysarthric speech recognition,” inICASSP, 2019, pp. 5836–5840

    2019

  12. [21]

    Other approaches include fine-tuning speaker-independent models on individual speakers [24–26] and leveraging latent features [27–30]

    and x-vectors [22] are used as auxiliary features for self- supervised models like wav2vec 2.0 [23] to capture speaker- specific characteristics. Other approaches include fine-tuning speaker-independent models on individual speakers [24–26] and leveraging latent features [27–3...

    2025 arXiv

  13. [22]

    SpecAugment: A simple data augmentation method for auto- matic speech recognition,

    D. S. Park, W. Chan, Y . Zhang, C.-C. Chiu, B. Zoph, E. D. Cubuk, and Q. V . Le, “SpecAugment: A simple data augmentation method for auto- matic speech recognition,” inInterspeech, 2019

    2019

  14. [23]

    DuTa-VC: A Duration-aware Typical-to-atypical V oice Conversion Approach with Diffusion Probabilistic Model,

    H. Wang, T. Thebaud, J. Villalba, M. Sydnor, B. Lammers, N. Dehak, and L. Moro-Velazquez, “DuTa-VC: A Duration-aware Typical-to-atypical V oice Conversion Approach with Diffusion Probabilistic Model,” inInter- speech, 2023, pp. 1548–1552

    2023

  15. [24]

    Syn- thesizing dysarthric speech using multi-speaker TTS for dysarthric speech recognition,

    M. Soleymanpour, M. T. Johnson, R. Soleymanpour, and J. Berry, “Syn- thesizing dysarthric speech using multi-speaker TTS for dysarthric speech recognition,” inICASSP, 2022, pp. 7382–7386

    2022

  16. [25]

    Synthesis of new words for improved dysarthric speech recognition on an expanded vocabu- lary,

    J. Harvill, D. Issa, M. Hasegawa-Johnson, and C. Yoo, “Synthesis of new words for improved dysarthric speech recognition on an expanded vocabu- lary,” inICASSP, 2021, pp. 6428–6432

    2021

  17. [26]

    Speaker adaptation for wav2vec2 based dysarthric ASR,

    M. K. Baskar, T. Herzig, D. Nguyen, M. Diez, T. Polzehl, L. Burget, and J. ˇCernock´y, “Speaker adaptation for wav2vec2 based dysarthric ASR,” in Interspeech, 2022, pp. 3403–3407

    2022

  18. [27]

    Maximum likelihood linear transformations for HMM-based speech recognition,

    M. Gales, “Maximum likelihood linear transformations for HMM-based speech recognition,”Computer Speech & Language, vol. 12, no. 2, pp. 75– 98, 1998

    1998

  19. [28]

    X- Vectors: Robust DNN Embeddings for Speaker Recognition,

    D. Snyder, D. Garcia-Romero, G. Sell, D. Povey, and S. Khudanpur, “X- Vectors: Robust DNN Embeddings for Speaker Recognition,” inICASSP, 2018, pp. 5329–5333

    2018

  20. [29]

    wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations,

    A. Baevski, Y . Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations,” in NeurIPS, 2020, pp. 12 449–12 460

    2020

  21. [30]

    Two-step acoustic model adap- tation for dysarthric speech recognition,

    R. Takashima, T. Takiguchi, and Y . Ariki, “Two-step acoustic model adap- tation for dysarthric speech recognition,” inICASSP, 2020, pp. 6104–6108

    2020

  22. [31]

    An analysis of degenerating speech due to progressive dysarthria on asr performance,

    K. Tomanek, K. Seaver, P.-P. Jiang, R. Cave, L. Harrell, and J. R. Green, “An analysis of degenerating speech due to progressive dysarthria on asr performance,” inICASSP, 2023, pp. 1–5

    2023

  23. [32]

    Personalizing ASR for dysarthric and accented speech with limited data,

    J. Shoret al., “Personalizing ASR for dysarthric and accented speech with limited data,” inInterspeech, 2019, pp. 784–788

    2019

  24. [33]

    On-the-fly feature based rapid speaker adaptation for dysarthric and elderly speech recognition,

    M. Genget al., “On-the-fly feature based rapid speaker adaptation for dysarthric and elderly speech recognition,” inInterspeech, 2023

    2023

  25. [34]

    Enhancing dysarthric speech recognition for unseen speakers via prototype-based adaptation,

    S. Wang, S. Zhao, J. Zhou, A. Kong, and Y . Qin, “Enhancing dysarthric speech recognition for unseen speakers via prototype-based adaptation,” in Interspeech, 2024, pp. 1305–1309

    2024

  26. [35]

    Cross-lingual self-supervised speech representations for improved dysarthric speech recognition,

    A. Hernandez, P. A. P ´erez-Toro, E. N¨oth, J. R. Orozco-Arroyave, A. Maier, and S. H. Yang, “Cross-lingual self-supervised speech representations for improved dysarthric speech recognition,” inInterspeech, 2022

    2022

  27. [36]

    Self-supervised asr models and features for dysarthric and elderly speech recognition,

    S. Hu, X. Xie, M. Geng, Z. Jin, J. Deng, G. Li, Y . Wang, M. Cui, T. Wang, H. Meng, and X. Liu, “Self-supervised asr models and features for dysarthric and elderly speech recognition,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 3561–3575, 2024

    2024

  28. [37]

    Robust cross-etiology and speaker- independent dysarthric speech recognition,

    S. Singh, Q. Wang, Z. Zhong, C. Mendes, M. Hasegawa-Johnson, W. Abdulla, and S. R. Shahamiri, “Robust cross-etiology and speaker- independent dysarthric speech recognition,” 2025. [Online]. Available: https://arxiv.org/abs/2501.14994

    2025 arXiv

  29. [38]

    Fine-tuning automatic speech recognition for people with parkinson’s: An effective strategy for enhancing speech technology accessibility,

    X. Zheng, B. Phukon, and M. Hasegawa-Johnson, “Fine-tuning automatic speech recognition for people with parkinson’s: An effective strategy for enhancing speech technology accessibility,” inInterspeech, 2024

    2024

  30. [39]

    Personalizing large sequence-to-sequence speech foundation models with speaker representations,

    D. Wagner, I. Baumann, T. Ranzenberger, K. Riedhammer, and T. Bocklet, “Personalizing large sequence-to-sequence speech foundation models with speaker representations,” inSLT, 2024

    2024

  31. [40]

    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,” 2022. [Online]. Available: https://arxiv.org/abs/2212.04356

    2022 arXiv

  32. [41]

    Natural language guidance of high-fidelity text-to-speech with synthetic annotations,

    D. Lyth and S. King, “Natural language guidance of high-fidelity text-to-speech with synthetic annotations,” 2024. [Online]. Available: https://arxiv.org/abs/2402.01912

    2024 arXiv

  33. [42]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,”NeurIPS, 2017

    2017

  34. [43]

    Community-supported shared infrastructure in support of speech accessibility,

    M. Hasegawa-Johnsonet al., “Community-supported shared infrastructure in support of speech accessibility,”Journal of Speech, Language, and Hear- ing Research, vol. 67, no. 11, 2024

    2024

  35. [44]

    LoRA: Low-Rank Adaptation of Large Language Models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-Rank Adaptation of Large Language Models,” in ICLR, 2022

    2022

  36. [45]

    Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning ,

    Q. Zhang, M. Chen, A. Bukharin, P. He, Y . Cheng, W. Chen, and T. Zhao, “Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning ,” in ICLR, 2023

    2023

  37. [46]

    Parler-TTS,

    Y . Lacombe, V . Srivastav, and S. Gandhi, “Parler-TTS,” 2024. [Online]. Available: https://github.com/huggingface/parler-tts

    2024

  38. [47]

    Simple and controllable music generation,

    J. Copet, F. Kreuk, I. Gat, T. Remez, D. Kant, G. Synnaeve, Y . Adi, and A. Defossez, “Simple and controllable music generation,” inNeurIPS, vol. 36, 2023, pp. 47 704–47 720

    2023

  39. [48]

    Scaling instruction-finetuned language models,

    H. W. Chunget al., “Scaling instruction-finetuned language models,”

  40. [49]

    BERTScore: Evaluating text generation with BERT,

    T. Zhang, V . Kishore, F. Wu, K. Q. Weinberger, and Y . Artzi, “BERTScore: Evaluating text generation with BERT,” inICLR, 2020

    2020

  41. [50]

    High- fidelity audio compression with improved RVQGAN,

    R. Kumar, P. Seetharaman, A. Luebs, I. Kumar, and K. Kumar, “High- fidelity audio compression with improved RVQGAN,” inNeurIPS, 2023

    2023

  42. [51]

    High fidelity neural audio compression,

    A. D ´efossez, J. Copet, G. Synnaeve, and Y . Adi, “High fidelity neural audio compression,”TMLR, 2023

    2023

  43. [52]

    Phi-3 technical report: A highly capable language model locally on your phone,

    M. Abdinet al., “Phi-3 technical report: A highly capable language model locally on your phone,” 2024. [Online]. Available: https: //arxiv.org/abs/2404.14219

    2024 arXiv

  44. [53]

    The Llama 3 herd of models,

    A. Grattafioriet al., “The Llama 3 herd of models,” 2024. [Online]. Available: https://arxiv.org/abs/2407.21783

    2024 arXiv

  45. [54]

    Textbooks are all you need II: phi-1.5 technical report,

    Y . Li, S. Bubeck, R. Eldan, A. D. Giorno, S. Gunasekar, and Y . T. Lee, “Textbooks are all you need II: phi-1.5 technical report,” 2023. [Online]. Available: https://arxiv.org/abs/2309.05463

    2023 arXiv

  46. [55]

    ECAPA-TDNN: Em- phasized channel attention, propagation and aggregation in tdnn based speaker verification,

    B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA-TDNN: Em- phasized channel attention, propagation and aggregation in tdnn based speaker verification,” inInterspeech, 2020, pp. 3830–3834

    2020

  47. [56]

    Unveiling interpretability in self-supervised speech representations for parkinson’s diagnosis,

    D. Gimeno-G ´omez, C. Botelho, A. Pompili, A. Abad, and C.-D. Mart´ınez-Hinarejos, “Unveiling interpretability in self-supervised speech representations for parkinson’s diagnosis,” 2024. [Online]. Available: https://arxiv.org/abs/2412.02006

    2024 arXiv

  48. [57]

    MENLI: Robust evaluation metrics from natural lan- guage inference,

    Y . Chen and S. Eger, “MENLI: Robust evaluation metrics from natural lan- guage inference,”Transactions of the Association for Computational Lin- guistics, vol. 11, pp. 804–825, 2023

    2023

  49. [58]

    V oxCeleb: A Large-Scale Speaker Identification Dataset,

    A. Nagrani, J. S. Chung, and A. Zisserman, “V oxCeleb: A Large-Scale Speaker Identification Dataset,” inInterspeech, 2017

    2017

  50. [59]

    V oxceleb2: Deep speaker recognition,

    J. S. Chung, A. Nagrani, and A. Zisserman, “V oxceleb2: Deep speaker recognition,” inInterspeech, 2018, pp. 1086–1090

    2018

  51. [60]

    HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units,

    W. Hsu, B. Bolte, Y . Tsai, K. Lakhotia, R. Salakhutdinov, and A. Mohamed, “HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units,”IEEE TASLP, 2021

    2021

  52. [61]

    Framewise phoneme classification with bidirectional lstm networks,

    A. Graves and J. Schmidhuber, “Framewise phoneme classification with bidirectional lstm networks,” inProceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005., 2005, pp. 2047–2052

    2005

  53. [62]

    Connectionist temporal classification: labelling unsegmented sequence data with recur- rent neural networks,

    A. Graves, S. Fern ´andez, F. Gomez, and J. Schmidhuber, “Connectionist temporal classification: labelling unsegmented sequence data with recur- rent neural networks,” inICML, 2006, p. 369–376

    2006

  54. [64]

    Large language models for dysfluency detection in stuttered speech,

    D. Wagner, S. P. Bayerl, I. Baumann, E. N ¨oth, K. Riedhammer, and T. Bocklet, “Large language models for dysfluency detection in stuttered speech,” inInterspeech, 2024, pp. 5118–5122

    2024

  55. [65]

    Layer-wise analysis of a self- supervised speech representation model,

    A. Pasad, J.-C. Chou, and K. Livescu, “Layer-wise analysis of a self- supervised speech representation model,” inASRU, 2021, pp. 914–921

    2021

  56. [66]

    Openvoice: Versatile instant voice cloning,

    Z. Qin, W. Zhao, X. Yu, and X. Sun, “Openvoice: Versatile instant voice cloning,” 2024. [Online]. Available: https://arxiv.org/abs/2312.01479

    2024 arXiv

  57. [2022]

    Available: https://arxiv.org/abs/2210.11416

    [Online]. Available: https://arxiv.org/abs/2210.11416

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.