Pith. sign in

REVIEW 4 major objections 5 minor 39 references

WHISTRESS: Enriching Transcriptions with Sentence Stress Detection

T0 review · 4 major / 5 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 A frozen transcription model can read sentence stress from its own internal embeddings, trained entirely on synthetic speech.

desk verdict Useful alignment-free stress head and a clever synthetic data pipeline, but the Expresso comparison that drives the 'outperforms' claim is internally inconsistent and needs fixing. read the letter →

arxiv 2505.19103 v1 pith:VAUQW3UP submitted 2025-05-25 cs.CL cs.SDeess.AS

Iddo Yosha , Dorin Shteyman , Yossi Adi This is my paper · ORCID
classification cs.CLcs.SDeess.AS
keywords sentencestressdetectionprosodyspeechemphasisautomaticrecognitionzero-shotgeneralizationsynthetictrainingdataalignment-freeWhisper
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

WHISTRESS attempts to show that sentence stress—which words in a spoken sentence carry emphasis—can be detected and added to a transcript without needing word timestamps, forced alignment, or any extra input at inference time. The proposed method attaches a small learned head to a frozen Whisper speech-recognition model, and the training signal comes entirely from TINYSTRESS-15K, a roughly 15-hour synthetic dataset built by having a language model mark stressed words in everyday sentences and a text-to-speech service render them with exaggerated prosody. The central claim is that this synthetic-only training transfers to natural human speech: WHISTRESS outperforms aligner-dependent baselines on its validation sets and shows zero-shot performance that beats an existing emphasis classifier on one benchmark and is competitive on another. If the claim holds, stress annotation becomes a cheap add-on for automatic speech recognition rather than a data-hungry specialist task.

What carries the argument

The load-bearing object is the stress-detection head: one additional Whisper decoder block that cross-attends over the encoder and decoder embeddings of the 9th backbone layer, followed by a two-layer fully connected binary classifier that scores each generated token. The 9th layer is chosen empirically because it balances prosodic detail and linguistic content better than earlier or later layers. The synthetic data pipeline is the other half: sentences from TinyStories, stress labels produced by a language model, and text-to-speech synthesis that modifies emphasized words with reduced speaking rate (30–85%), raised volume (+3–6 dB), and higher pitch (+1.5 semitones), with Gaussian noise added for variability.

What would settle it

Look for a result that would break the transfer claim: take a natural, multi-speaker corpus with expert linguistic annotations of sentence stress and run the synthetic-trained WHISTRESS zero-shot alongside an aligner-based baseline; if its F1 falls to near chance while the baseline stays strong, the text-to-speech-to-human transfer assumption fails. A complementary check is to test whether WHISTRESS only learns the exact SSML recipe—longer, louder, higher—by inspecting false positives on naturally stressed words that do not follow that recipe.

Watch

Extended reading notes

Core claim

The paper's core discovery is that a frozen speech-recognition model already stores enough prosodic information in its intermediate layers that a small decoder-style head can read out sentence stress per token. The head is trained with binary cross-entropy on token-level stress labels, after a relaxed word-length filter prevents Whisper transcription errors from shifting labels; no word timestamps are used in training, and none are needed at inference. The key reported findings are that WHISTRESS reaches 0.909 F1 on TINYSTRESS-15K and 0.961 F1 on Aix-MARSEC, beating baselines that receive true or force-aligned word boundaries, and that the version trained only on synthetic data reaches 0.689 F1 zero-shot on Expresso, ahead of EmphaClass's 0.654. On EmphAssess, zero-shot WHISTRESS reaches 0.797 F1 with high recall, while a version trained on about 1,500 in-domain samples reaches 0.943 F1, slightly above EmphaClass.

Load-bearing premise

The claim stands on the assumption that the synthetic text-to-speech renderings, with stressed words made longer, louder, and higher-pitched, capture enough of natural sentence stress that a model trained on them will transfer to real human speech.

Editorial extensions

If this is right

  • Transcription systems can enrich outputs with emphasis markup without changing the recognized words, since the backbone is frozen and the head is purely additive.
  • A fully automated pipeline can produce stress-annotated training data at scale, avoiding closed-source corpora and human annotators.
  • The alignment-free design means inference does not require forced alignment or timestamp estimation, removing an error source and a scalability bottleneck.
  • The layer analysis implies that prosody-aware heads should be attached at intermediate layers; the 9th layer outperforms the final layer for stress detection.
  • If the zero-shot numbers transfer, stress detection becomes feasible for domains where no in-domain stress annotations exist.

Reading between the lines

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

  • Editorial inference: the same frozen-backbone-plus-head recipe could plausibly be applied to other word-level paralinguistic labels, such as emotion or uncertainty, where synthetic data can be generated with controlled prosody; the paper only anticipates this possibility.
  • Editorial inference: the layer-9 sweet spot suggests a general design heuristic—intermediate layers of speech encoders may be better than final layers for prosody-related tasks—that could be tested on other languages and model sizes.
  • Editorial inference: the label-alignment trick could be strengthened by edit-distance matching over whole words instead of dropping samples with mismatched word counts, perhaps extending the method to noisier automatic speech recognition conditions.
  • Editorial inference: because the paper's Aix-MARSEC analysis shows most unstressed words are a small recurring set, part of the reported gain may come from lexical predictability rather than acoustic sensitivity; a synthetic dataset with more unpredictable unstressed words would test whether true prosodic transfer remains.
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 / 5 minor

Summary. The paper introduces WHISTRESS, an alignment-free extension of Whisper for token-level sentence stress detection. A stress head consisting of one Whisper decoder block and a two-layer FCNN is trained on a new synthetic dataset TINYSTRESS-15K (~15 hours), created by using GPT-4o-mini to label stressed words in TinyStories sentences and Google TTS with SSML prosodic modifications. The backbone Whisper is frozen. The paper reports that WHISTRESS outperforms a BLSTM acoustic baseline, CRF [11], hierarchical BLSTM [10], and EmphaClass [29] on Aix-MARSEC and Expresso, with zero-shot transfer from synthetic to natural speech. It also includes an analysis of Whisper layers' prosodic information and a layer-selection study.

Significance. WHISTRESS addresses a real gap: existing sentence stress detectors require forced alignment or manual annotations, while this model needs no word timestamps at inference. The synthetic data pipeline is a practical contribution, and the authors release code, weights, and dataset. The layer analysis (Section 4) is informative and connects to known acoustic/semantic trade-offs in self-supervised speech models. The main empirical claim—strong zero-shot generalization to natural benchmarks—would be significant if confirmed, since it would show that synthetic prosodic training transfers across speakers and recording conditions.

major comments (4)
  1. [§5.1 vs Table 2] The evaluation protocol for Expresso is inconsistent. Section 5.1 states 'we use the same test set configuration as [29], selecting the speakers with IDs ex01 and ex02,' but the caption of Table 2 says the 0-shot variant is evaluated 'over all four speakers,' with only the in-domain variant (marked *) evaluated on ex01/ex02. If the WHISTRESS[0-shot] Expresso row (F1 0.689) includes ex03/ex04, it is not directly comparable to EmphaClass's 0.654, which was computed on ex01/ex02. The claim that WHISTRESS 'surpasses EmphaClass on Expresso' therefore rests on an ambiguous split. The same ambiguity applies to EmphAssess: the 0-shot row may include all four voices, while EmphaClass's F1 is presumably on ex01/ex02. Please report the exact speaker split for every row, provide per-speaker results, and re-compute the comparison on the same subset.
  2. [§5 / Table 2] No confidence intervals, significance tests, or multiple-seed results are reported. The margins that support the abstract's comparative claim are small (e.g., WHISTRESS 0.689 vs EmphaClass 0.654 on Expresso; 0.943 vs 0.938 on EmphAssess after in-domain training). Given that training involves random initialization and stochastic optimization, the reported F1 differences may not be stable. Please provide standard errors or confidence intervals over several seeds for the key comparisons, or at least state the number of runs.
  3. [Abstract / §5.3] The abstract states that WHISTRESS 'outperforms existing methods,' but the results in Table 2 show that WHISTRESS[0-shot] on EmphAssess (F1 0.797) is substantially below EmphaClass (F1 0.938). The paper's discussion acknowledges that EmphaClass was trained on a closed-source, domain-matched dataset, but the unqualified abstract claim is too strong. Please qualify the claim to the specific benchmarks and settings where the improvement holds, or add a direct comparison controlling for training data.
  4. [§2 / §5.3 (TINYSTRESS-15K validation)] The reported validation on TINYSTRESS-15K is partially circular: the test labels are produced by the same automated pipeline (GPT-4o-mini selection plus SSML-modified TTS) that generated the training data. This makes the test set useful for model selection but not for measuring absolute accuracy on natural speech. The external zero-shot evaluations on Aix-MARSEC, Expresso, and EmphAssess are independent and should remain the primary evidence; please explicitly state this limitation in Section 5.3 or Section 2. The choice of layer 9 (Table 1) is also made on this same test set, so the circularity may influence architecture decisions.
minor comments (5)
  1. [Throughout] The paper uses 'W HISTRESS' and 'T INY STRESS -15K' with unusual spacing in several places (e.g., Abstract, Sections 2 and 3); the dataset name is written 'TINYSTRESS-15K' in the abstract and title but 'T INY STRESS -15K' in the body. Please normalize the notation.
  2. [Table 2] The EmphAssess WHISTRESS row has concatenated values '0.945*0.942*0.943*' that are hard to read; use proper column separators or formatting for the starred entries.
  3. [References] References [9] and [29] are the same work (de Seyssel et al., 'Emphassess: a prosodic benchmark on assessing emphasis transfer in speech-to-speech models'); please consolidate them into a single reference.
  4. [§4.1] The size and composition of the CREMA-D subset used for the prosodic information analysis are not specified; adding this detail would improve reproducibility.
  5. [§5.2] The baseline feature extraction is described as 'duration, mean energy and max pitch' inspired by [10], but the exact windowing, aggregation, and alignment procedure are not specified, making the baseline difficult to reproduce.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; central zero-shot claims rest on external benchmarks.

full rationale

The paper's derivation chain is not circular. WHISTRESS is trained on synthetic TINYSTRESS-15K, where stress labels are produced by GPT-4o-mini and realized through SSML prosody modifications; the model is then evaluated zero-shot on independently annotated or externally generated corpora (Aix-MARSEC, Expresso, EmphAssess). None of the reported external evaluations is defined in terms of WHISTRESS's own outputs or fitted parameters, and no self-citation supplies the load-bearing premise. The TINYSTRESS-15K validation row is an in-distribution check, but the paper explicitly separates it from zero-shot generalization claims. The choice of Whisper layer 9 is a model-selection decision on the synthetic validation set, not a parameter fitted to the external benchmark labels. The Expresso/EmphAssess comparison exhibits a possible test-set inconsistency (Table 2 caption says all four speakers for 0-shot, while Section 5.1 says ex01/ex02 used for fair comparison), but this is a measurement/fairness issue, not a circular reduction of the predicted quantity to its inputs. Therefore no specific circular step can be quoted, and the circularity score is 0.

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

The central claim rests on the synthetic data faithfully encoding natural stress cues and on Whisper's pretrained features being sufficient. The prosodic adjustment settings and the selected layer are hand-chosen parameters rather than fitted values. No new physical or scientific entities are introduced.

free parameters (4)
  • SSML prosodic adjustments = duration -30% to -85%, volume +3 to +6 dB, pitch +1.5 semitones
    Set by hand based on [4] and manual refinement to make synthetic speech sound natural; determines the acoustic cues in all training data.
  • Whisper layer for stress head input = 9th layer
    Selected via validation on TINYSTRESS-15K (Table 1); used for all reported stress detection results.
  • Training epochs = 4 epochs (2 epochs for zero-shot variant)
    Heuristic choice; no early stopping or hyperparameter tuning described.
  • Word-level stress threshold = word stressed if at least one token predicted stressed
    Conversion rule for evaluating word-level F1 rather than token-level.
assumptions (4)
  • domain assumption Whisper's frozen representations contain sufficient acoustic and linguistic information for sentence stress detection.
    Used throughout Section 3; the entire method depends on these features being predictive.
  • domain assumption Google TTS speech with manually tuned SSML prosody is a valid proxy for natural expressive speech.
    Central to Section 2; transfer to real benchmarks is expected despite synthetic training.
  • domain assumption GPT-4o-mini's stress label selections reflect natural sentence stress.
    Used in Section 2 to create all training and test labels for TINYSTRESS-15K; no human validation is reported.
  • domain assumption Binary per-word stress labels adequately represent sentence stress.
    Adopted in Section 1 and throughout; ignores graded prominence and multi-word emphasis patterns.

how reviews work

0 comments
Cite this review

Pith. "Pith review of WHISTRESS: Enriching Transcriptions with Sentence Stress Detection." pith.science (2026). https://pith.science/paper/VAUQW3UP

@misc{pith2026250519103,
  author       = {Pith},
  title        = {Pith review of: WHISTRESS: Enriching Transcriptions with Sentence Stress Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VAUQW3UP}},
  note         = {Machine review of arXiv:2505.19103}
}
read the original abstract

Spoken language conveys meaning not only through words but also through intonation, emotion, and emphasis. Sentence stress, the emphasis placed on specific words within a sentence, is crucial for conveying speaker intent and has been extensively studied in linguistics. In this work, we introduce WHISTRESS, an alignment-free approach for enhancing transcription systems with sentence stress detection. To support this task, we propose TINYSTRESS-15K, a scalable, synthetic training data for the task of sentence stress detection which resulted from a fully automated dataset creation process. We train WHISTRESS on TINYSTRESS-15K and evaluate it against several competitive baselines. Our results show that WHISTRESS outperforms existing methods while requiring no additional input priors during training or inference. Notably, despite being trained on synthetic data, WHISTRESS demonstrates strong zero-shot generalization across diverse benchmarks. Project page: https://pages.cs.huji.ac.il/adiyoss-lab/whistress.

Figures

Figures reproduced from arXiv: 2505.19103 by the authors.

Figure 1
Figure 1. WHISTRESS Architecture. The Whisper model is kept frozen during training. The extension is a transformer decoder block with cross-attention for the audio embeddings, followed by an FCNN classifier that outputs the stress score per token. suggested fine-tuning Whisper [14] by adding tokens in the text for stress labeling. However, their method do not take into account possible performance degradation caused by alteri… view at source ↗
Figure 2
Figure 2. Prosodic features prediction by Mean Absolute Error (MAE) percentage of Whisper layer embeddings. A lower MAE percentage indicates better prediction. Each curve shows con￾fidence intervals. a pre-processing step before training, and is alignment-free dur￾ing inference. Notably, at any stage, it does not require word￾level timestamps (i.e., no time alignment is needed), as Whisper inherently aligns generated tokens w… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 23 canonical work pages

  1. [11]

    Van Heuven,Acoustic Correlates and Perceptual Cues of Word and Sentence Stress: Theories, Methods and Data , 12 2018, pp

    V . Van Heuven,Acoustic Correlates and Perceptual Cues of Word and Sentence Stress: Theories, Methods and Data , 12 2018, pp. 15–59

    work page 2018

  2. [10]

    Accent is predictable (if you’re a mind-reader),

    D. Bolinger, “Accent is predictable (if you’re a mind-reader),” Language, vol. 48, no. 3, pp. 633–644, 1972. [Online]. Available: http://www.jstor.org/stable/412039

    work page 1972

  3. [29]

    Aix-MARSEC is a speech corpus containing over 5 hours of 1980s BBC radio recordings, featuring 53 speakers across dif- ferent speech styles [27]

    benchmarks for evaluation, described below. Aix-MARSEC is a speech corpus containing over 5 hours of 1980s BBC radio recordings, featuring 53 speakers across dif- ferent speech styles [27]. In accordance with previous work [11], we define a word as stressed for the Aix-MARSEC ground-truth sentence-stress labels if it contains the first syl- lable in Jasse...

    work page

  4. [1]

    The linguistic formulation of stress falls into two perspectives, as described in [1]

    Introduction Theoretical work on sentence stress can be divided into linguis- tic and acoustic research. The linguistic formulation of stress falls into two perspectives, as described in [1]. The first, de- fines normal stress as a default pattern, independent of mean- ing, that follows phonological constraints [2]. The second per- spective views sentence...

    arXiv 2025

  5. [2]

    Transcription Selection

    Synthetic Data We develop an automated pipeline for synthetic dataset gener- ation, consisting of three key steps: (i) transcription selection, (ii) stressed words labeling, and (iii) speech synthesis. Transcription Selection. To get transcriptions of coherent, diverse, everyday language, we use the TinyStories dataset, a common choice for small language ...

    work page

  6. [3]

    Specifically, we propose generating two outputs for each input speech signal: the tran- scription and an emphasis score for each token in the transcrip- tion

    Method WHISTRESS enhances transcription models with additional sentence stress detection objective. Specifically, we propose generating two outputs for each input speech signal: the tran- scription and an emphasis score for each token in the transcrip- tion. To achieve this, we modify both the architecture of the Whisper model and its training procedure. ...

    work page

  7. [4]

    Furthermore, we investigate how WHISTRESS identifies sentence stress by evaluating different layers as input for the stress detection head

    Model Analysis We analyze Whisper’s internal representations to determine which layers capture prosodic features, specifically pitch, energy, and duration. Furthermore, we investigate how WHISTRESS identifies sentence stress by evaluating different layers as input for the stress detection head. 4.1. Prosodic information analysis To understand where prosod...

    work page

  8. [5]

    Our experiments across different datasets validate our model’s ability to identify stressed words accurately while also demonstrating impressive generalization capabilities

    Results We evaluate our model by comparing it to previously proposed approaches for sentence stress detection and established bench- marks. Our experiments across different datasets validate our model’s ability to identify stressed words accurately while also demonstrating impressive generalization capabilities. We re- port performance using standard clas...

    work page

Show all 39 references
  1. [6]

    We also presented a fully automated pipeline for synthetically generating training data specific for sentence stress detection

    Conclusion In this work, we introduced W HISTRESS , an extension to the Whisper model to enrich its transcriptions by marking sentence stress while preserving its core functionality. We also presented a fully automated pipeline for synthetically generating training data specif...

  2. [7]

    Word prominence de- tection using robust yet simple prosodic features,

    T. Mishra, V . R. Sridhar, and A. Conkie, “Word prominence de- tection using robust yet simple prosodic features,” in Interspeech 2012, 2012, pp. 1864–1867

    2012

  3. [8]

    D. R. Ladd, Intonational Phonology , 2nd ed., ser. Cambridge Studies in Linguistics. Cambridge University Press, 2008

    2008

  4. [9]

    The sound pattern of english,

    N. Chomsky and M. Halle, “The sound pattern of english,” 1968. [Online]. Available: https://api.semanticscholar.org/CorpusID: 60457972

    1968

  5. [12]

    Prosodic stress revisited: Reassess- ing the role of fundamental frequency,

    R. Silipo and S. Greenberg, “Prosodic stress revisited: Reassess- ing the role of fundamental frequency,” 01 2000

    2000

  6. [13]

    3pro – an unsupervised method for the automatic detection of sentence prominence in speech,

    S. Kakouros and O. R ¨as¨anen, “3pro – an unsupervised method for the automatic detection of sentence prominence in speech,” Speech Communication, vol. 82, p. 67–84, 09 2016

    2016

  7. [14]

    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

  8. [15]

    Crowdsourced and automatic speech prominence estimation,

    M. Morrison, P. Pawar, N. Pruyne, J. Cole, and B. Pardo, “Crowdsourced and automatic speech prominence estimation,”

  9. [16]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” 2019. [Online]. Available: https://arxiv.org/abs/ 1810.04805

    2019 arXiv

  10. [18]

    Joint detection of sentence stress and phrase boundary for prosody,

    B. Lin, L. Wang, X. Feng, and J. Zhang, “Joint detection of sentence stress and phrase boundary for prosody,” in Interspeech, 2020. [Online]. Available: https://api.semanticscholar.org/CorpusID:226201825

    2020

  11. [19]

    Automatic sentence stress feedback for non-native english learners,

    G. Lee, H.-Y . Lee, J. Song, B. Kim, S. Kang, J. Lee, and H. Hwang, “Automatic sentence stress feedback for non-native english learners,” Computer Speech & Language , vol. 41, 06 2016

    2016

  12. [20]

    Hierarchical representa- tion of prosody for statistical speech synthesis,

    A. Suni, D. Aalto, and M. Vainio, “Hierarchical representa- tion of prosody for statistical speech synthesis,” arXiv preprint arXiv:1510.01949, 2015

    2015 arXiv

  13. [21]

    Non-verbal information in spontaneous speech-towards a new framework of analysis,

    T. Biron, M. Barboy, E. Ben-Artzy, A. Golubchik, Y . Marmor, S. Szekely, Y . Winter, and D. Harel, “Non-verbal information in spontaneous speech-towards a new framework of analysis,”arXiv preprint arXiv:2403.03522, 2024

    2024 arXiv

  14. [22]

    Whisperx: Time-accurate speech transcription of long-form audio,

    M. Bain, J. Huh, T. Han, and A. Zisserman, “Whisperx: Time-accurate speech transcription of long-form audio,” 2023. [Online]. Available: https://arxiv.org/abs/2303.00747

    2023 arXiv

  15. [23]

    Fine-tuning is fine, if calibrated,

    Z. Mai, A. Chowdhury, P. Zhang, C.-H. Tu, H.-Y . Chen, V . Pahuja, T. Berger-Wolf, S. Gao, C. Stewart, Y . Su, and W.-L. Chao, “Fine-tuning is fine, if calibrated,” 2024. [Online]. Available: https://arxiv.org/abs/2409.16223

    2024 arXiv

  16. [24]

    An introduction to the boot- strap,

    B. Efron and R. Tibshirani, “An introduction to the boot- strap,” 1995. [Online]. Available: https://api.semanticscholar.org/ CorpusID:123398593

    1995

  17. [25]

    Bert, can he predict contrastive focus? predicting and controlling prominence in neural tts using a language model,

    B. Stephenson, L. Besacier, L. Girin, and T. Hueber, “Bert, can he predict contrastive focus? predicting and controlling prominence in neural tts using a language model,” 2022. [Online]. Available: https://arxiv.org/abs/2207.01718

    2022 arXiv

  18. [26]

    Tinystories: How small can language models be and still speak coherent english?

    R. Eldan and Y . Li, “Tinystories: How small can language models be and still speak coherent english?” 2023. [Online]. Available: https://arxiv.org/abs/2305.07759

    2023 arXiv

  19. [27]

    S. Bird, E. Klein, and E. Loper, Natural Language Processing with Python, 1st ed. O’Reilly Media, Inc., 2009

    2009

  20. [28]

    Gpt-4o system card,

    A. Hurst, A. Lerer, A. P. Goucher, A. Perelman, A. Ramesh, A. Clark, A. Ostrow, A. Welihinda, A. Hayes, A. Radford et al., “Gpt-4o system card,” arXiv preprint arXiv:2410.21276, 2024

    2024 arXiv

  21. [30]

    Crema-d: Crowd-sourced emotional multimodal actors dataset,

    H. Cao, D. G. Cooper, M. K. Keutmann, R. C. Gur, A. Nenkova, and R. Verma, “Crema-d: Crowd-sourced emotional multimodal actors dataset,”IEEE Transactions on Affective Computing, vol. 5, no. 4, pp. 377–390, 2014

    2014

  22. [31]

    Random forests,

    L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001

    2001

  23. [32]

    Montreal Forced Aligner: Trainable Text-Speech Alignment Using Kaldi,

    M. McAuliffe, M. Socolof, S. Mihuc, M. Wagner, and M. Son- deregger, “Montreal Forced Aligner: Trainable Text-Speech Alignment Using Kaldi,” in Proc. Interspeech 2017 , 2017, pp. 498–502

    2017

  24. [33]

    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,” in2021 IEEE Auto- matic Speech Recognition and Understanding Workshop (ASRU). IEEE, 2021, pp. 914–921

    2021

  25. [34]

    wav2vec 2.0: A framework for self-supervised learning of speech repre- sentations,

    A. Baevski, Y . Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A framework for self-supervised learning of speech repre- sentations,” Advances in neural information processing systems , vol. 33, pp. 12 449–12 460, 2020

    2020

  26. [35]

    The aix-marsec project: an evolutive database of spoken british english,

    C. Auran, C. Bouzon, and D. J. Hirst, “The aix-marsec project: an evolutive database of spoken british english,” Speech Prosody 2004 , 2004. [Online]. Available: https: //api.semanticscholar.org/CorpusID:204957167

    2004

  27. [36]

    Ex- presso: A benchmark and analysis of discrete expressive speech resynthesis,

    T. A. Nguyen, W.-N. Hsu, A. d’Avirro, B. Shi, I. Gat, M. Fazel- Zarani, T. Remez, J. Copet, G. Synnaeve, M. Hassid et al., “Ex- presso: A benchmark and analysis of discrete expressive speech resynthesis,” arXiv preprint arXiv:2308.05725, 2023

    2023 arXiv

  28. [37]

    Em- phassess: a prosodic benchmark on assessing emphasis transfer in speech-to-speech models,

    M. de Seyssel, A. D’Avirro, A. Williams, and E. Dupoux, “Em- phassess: a prosodic benchmark on assessing emphasis transfer in speech-to-speech models,” arXiv preprint arXiv:2312.14069 , 2023

    2023 arXiv

  29. [38]

    Stress in modern english,

    W. Jassem, “Stress in modern english,” Bulletin de la Soci´et´e Lin- guistique Polonaise, vol. 11, pp. 23–49, 1952

    1952

  30. [39]

    Bidirectional lstm-crf models for sequence tagging,

    Z. Huang, W. Xu, and K. Yu, “Bidirectional lstm-crf models for sequence tagging,” 2015. [Online]. Available: https://arxiv.org/abs/1508.01991

    2015 arXiv

  31. [2023]

    Available: https://arxiv.org/abs/2310.08464

    [Online]. Available: https://arxiv.org/abs/2310.08464

    arXiv

Pith tools

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