Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Should Top-Down Clustering Affect Boundaries in Unsupervised Word Discovery?

T0 review · 3 major / 6 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 A simple bottom-up pipeline matches a top-down system for unsupervised word discovery on five languages while running about five times faster, and the bottleneck is the clustering step, not boundary detection.

desk verdict Honest empirical study with a strong oracle-boundary diagnostic, but the bottom-up vs top-down head-to-head is not fully controlled because the two systems use different prominence hyperparameters. read the letter →

arxiv 2507.19204 v2 pith:YEZANSJB submitted 2025-07-25 eess.AS cs.CLcs.SD

Simon Malan , Benjamin van Niekerk , Herman Kamper This is my paper · ORCID
classification eess.AScs.CLcs.SD
keywords unsupervisedworddiscoverysegmentationlexiconlearningzero-resourcespeechself-supervisedrepresentationsHuBERTES-KMeansprominence-basedboundaries
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 asks whether top-down information — feeding clustering results back into boundary selection — is actually necessary for full-coverage unsupervised word discovery from raw speech. It builds two systems that share the same HuBERT features and K-means lexicon recipe: a simple bottom-up pipeline that fixes word boundaries at prominence peaks in adjacent-frame dissimilarity and then clusters the resulting segments, and ES-KMeans+, a modernized dynamic-programming method that iteratively reselects boundaries to fit the current K-means model. On the five-language ZeroSpeech Track 2 benchmark, both reach comparable state-of-the-art results, with the bottom-up system running about five times faster. Experiments that inject ground-truth boundaries show that even with perfect boundaries the lexicon has a normalized edit distance around 30 percent, indicating that the clustering step, not boundary refinement, is the main bottleneck. The paper therefore recommends focusing future work on better clustering and more discriminative word-like representations.

What carries the argument

The key machinery is a controlled two-system comparison plus oracle-boundary probes that separate segmentation quality from clustering quality. Both systems encode speech with an intermediate HuBERT layer [28]. The bottom-up system uses a previously proposed prominence-based detector [20]: it computes the cosine distance between adjacent frames, smooths the dissimilarity curve, and marks word boundaries at peaks above a prominence threshold. Each segment is embedded by averaging its PCA-reduced HuBERT features and normalizing to the unit sphere, and K-means on the resulting fixed-dimensional vectors builds the lexicon [30]. ES-KMeans+ reuses the same features, embeddings, and clustering but replaces the fixed boundaries with a Viterbi dynamic-programming search over a candidate boundary set: it selects the segmentation that minimizes the sum of length-weighted distances from each segment embedding to its nearest cluster centroid [19]. Feeding both methods candidate boundary sets of different granularities—prominence, ground-truth phones, syllables, and words—is the probe that isolates the lexicon-building step as the bottleneck.

What would settle it

Repeat the oracle-boundary experiment with perfect word boundaries but replace K-means with an oracle clustering that assigns each segment to its ground-truth type before computing NED; if NED stays near 31 percent, the averaged embeddings are the limiting factor rather than the clustering algorithm.

Watch

Extended reading notes

Core claim

The central claim is that top-down clustering feedback is not needed to get competitive full-coverage word discovery: a bottom-up pipeline of prominence-based boundary detection on HuBERT features followed by K-means on averaged, unit-normalized segment embeddings matches or nearly matches the top-down ES-KMeans+ on English, French, Mandarin, German, and Wolof, while being roughly five times faster. The top-down method is valuable only in a specific regime: when the candidate boundary set over-segments relative to true words, ES-KMeans+ trades boundary recall for precision and improves segmentation scores without hurting lexicon quality. The decisive evidence is the oracle experiment: with ground-truth word boundaries given to ES-KMeans+, the lexicon's normalized edit distance remains near 30 percent, and moving from perfect boundaries to the system's own boundaries barely changes NED. From this the paper concludes that the lexicon-building clustering step is the limiting factor for both methods, and that improved clustering techniques and more discriminative word-like speech representations should be the priority.

Load-bearing premise

The conclusion that clustering is the bottleneck assumes that the averaged, unit-normalized HuBERT embeddings preserve enough word identity, since the oracle-boundary experiment cannot tell poor clustering apart from poor embeddings.

Editorial extensions

If this is right

  • Improved clustering and more discriminative word embeddings should yield larger gains in unsupervised word discovery than further boundary-detection work.
  • A bottom-up pipeline is a strong, cheap default; top-down refinement is only worth its added compute when the candidate boundary set clearly over-segments.
  • The oracle results imply that even perfect boundaries do not solve lexicon quality, so evaluations should report NED at fixed boundary quality to track progress on the representation side.
  • New top-down systems should configure their candidate boundary sets for high recall, since ES-KMeans+ reliably improves such sets by cutting spurious boundaries.

Reading between the lines

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

  • A direct test of the bottleneck claim would replace K-means with a clustering method that avoids over-clustering (for instance, agglomerative clustering with a learned distance) on the same averaged embeddings; if NED stays near 30 percent with perfect boundaries, the embeddings themselves are the limiting factor rather than the clusterer.
  • The averaging-and-normalizing embedding recipe may be casting away within-segment ordering information; sequence-level pooling such as attention pooling or a learned acoustic word embedding could close some of the NED gap the paper attributes to clustering.
  • The bottom-up system's fivefold speed advantage suggests it could serve as a candidate-boundary generator for ES-KMeans+, giving a practical way to get the top-down refinement benefit without paying its full runtime.
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

3 major / 6 minor

Summary. The paper asks whether top-down information from clustering should influence boundary selection in unsupervised word discovery. It introduces a simple bottom-up pipeline (prominence-based boundaries from HuBERT features, then K-means on averaged segment embeddings) and an updated top-down system, ES-KMeans+, which iteratively re-segments utterances using a K-means model. The two systems are compared on the five-language ZeroSpeech Track 2 benchmark and on LibriSpeech dev-clean. The main claims are that the two approaches perform comparably, that the bottom-up system is roughly five times faster, and that the lexicon-building component—rather than boundary refinement—is the main limiting factor. The paper also analyzes boundary types, discovered units, and the effect of candidate boundary sets, concluding that ES-KMeans+ is mainly useful when candidate boundaries over-segment and that future work should focus on better clustering and word-like representations.

Significance. If the central claims hold, the paper would provide a useful result for the zero-resource speech community: a lightweight bottom-up system can match the performance of a substantially more expensive top-down method, and effort is better spent on segment representations and clustering than on further boundary refinement. The paper's strengths include detailed ablations on LibriSpeech dev-clean (Table III), an oracle-boundary diagnostic (Table V), analysis of discovered cluster contents (Fig. 4), an open-source code release, and a clear qualitative distinction between word- and syllable-like boundary behavior. The recommendation to prioritize clustering and representations is actionable for the ZeroSpeech benchmark community. However, the experimental design does not fully control the comparison between the two systems, and the oracle experiments cannot separate the contributions of embeddings versus clustering to the reported bottleneck.

major comments (3)
  1. [Sec. V-C, Sec. VI-C, Table I] The claimed like-for-like comparison that isolates top-down clustering is not actually controlled. The bottom-up system uses a four-frame smoothing window with prominence threshold 0.75, while ES-KMeans+ is given a five-frame window with threshold 0.3 to create a 'high-recall' candidate set (Sec. V-C); the paper itself concedes in Sec. VI-C that 'their hyperparameter settings differ.' Because the two systems start from different boundary candidate sets, the ordering in Table I (ES-KMeans+ better NED in four languages, bottom-up better R-value in four languages) conflates the effect of top-down refinement with the effect of a different initial over-segmentation operating point. To support the central claim, the authors should run both methods on identical candidate boundary sets—e.g., both on the high-recall set and both on the directly tuned set—and show that the qualitative conclusion is unchanged.
  2. [Sec. VI-E, Table V, Sec. VII] The oracle-boundary experiment does not separate the quality of the segment embeddings from the quality of the clustering step, so it cannot directly support the claim that 'the clustering step is a limiting factor.' With ground-truth word boundaries, Prom. Seg. Clus. still has NED 31.1%, but the lexicon-building step includes both averaging/normalizing HuBERT features and K-means; a poor NED could be caused by the averaged embeddings not preserving word identity rather than by K-means over-clustering. The recommendation in Sec. VII to focus on 'other clustering methods' (point 2) is therefore not directly supported by this experiment. The authors should either isolate clustering quality (e.g., clustering oracle or stronger embeddings) or rephrase the conclusion to say that the combined representation-plus-clustering component is the bottleneck.
  3. [Sec. V-C, Table I, Table V] No variance information is reported despite ES-KMeans+ using random initialization (Sec. V-C) and the main comparative claims involving small metric differences (e.g., English NED 32.9 vs 33.5; French NED 43.2 vs 47.9). Without multiple seeds, error bars, or a paired significance test, the statements that the methods are 'comparable' and that ES-KMeans+ yields 'a slightly better lexicon' are not statistically grounded. Report at least five seeds for the main ZeroSpeech and LibriSpeech comparisons and state whether the observed differences are significant.
minor comments (6)
  1. [Abstract, Table II] The abstract's 'nearly five times faster' is not uniform across Table II: Wolof is 8/3 ≈ 2.7x, while German is 316/50 ≈ 6.3x. Consider reporting the range or saying 'typically four to five times faster.'
  2. [Sec. III] 'without loosing much phonetic information' contains a typo: 'loosing' should be 'losing.'
  3. [Sec. VI-E] The statement that 'ES-KMeans+ improves all scores except recall, increasing the R-value by 36.8% absolute in line 6' appears inconsistent with Table V: R-value in line 12 is -34.5 and in line 6 is 61.3, a difference of about 95.8 percentage points. Please check the number.
  4. [Sec. VI-D] The summary of where word remainders are clustered is not accompanied by a table or figure, making the claim difficult to verify; please illustrate the analysis or remove the summary.
  5. [Sec. V-B, Fig. 5] The bitrate metric is mentioned but never formally defined; a brief definition would help readers interpret Fig. 5.
  6. [Table II] The header 'Runtime (MIN)' should use 'min' or 'minutes' for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: core claims are benchmarked against implemented systems; minor hyperparameter confound affects experimental control, not derivation circularity.

full rationale

The paper's central comparisons are empirical and benchmark-based: the bottom-up prominence-segmentation-plus-clustering system and the updated ES-KMeans+ system are both implemented, run on ZeroSpeech Track 2 and LibriSpeech dev-clean, and reported directly in Tables I, II, IV, and V. No target metric is fitted into a derivation, and no claim reduces by construction to its inputs. The oracle-boundary experiments in Table V are an empirical upper-bound analysis: even with ground-truth word boundaries, NED is around 31%, which legitimately motivates the claim that the lexicon-building step (averaged HuBERT embeddings plus K-means) is a limiting factor. That experiment cannot separate embedding quality from clustering quality, but this is a limitation in attribution, not circularity. Several cited works are by the present authors (e.g., original ES-KMeans [19], DPDP [15], feature choices [39]), but the paper does not lean on those citations for its conclusions; it re-implements and ablates the components in Table III and reports its own numbers. The paper explicitly concedes that the two compared systems use different boundary-detection hyperparameters (Section V-C and VI-C), which is a real experimental confound for the claim that the comparison isolates top-down influence, but it is a matter of experimental control rather than a circular derivation. No self-definitional steps, fitted inputs renamed as predictions, load-bearing self-citations, imported uniqueness theorems, ansatz-smuggling citations, or renamings of known results were found.

Assumptions & free parameters 11 free parameters · 6 assumptions · 0 invented entities

No new physical or formal entities are introduced. The paper's conclusions are empirical and rest on benchmark assumptions and tuned hyperparameters; the largest free-parameter difference between the two compared systems is the boundary-detection configuration.

free parameters (11)
  • Prominence smoothing window (bottom-up) = 4 frames
    Tuned on development data to balance NED and R-value (Section V-C).
  • Prominence threshold (bottom-up) = 0.75
    Tuned on development data; controls which peaks become boundaries (Section V-C).
  • Prominence smoothing window (ES-KMeans+) = 5 frames
    Chosen for high-recall candidate boundaries so top-down refinement has boundaries to remove (Section V-C).
  • Prominence threshold (ES-KMeans+) = 0.3
    Chosen for high-recall candidate boundaries (Section V-C).
  • PCA dimensionality for clustering features = 250
    Reduction applied to HuBERT layer 12 features before averaging and clustering; based on development experiments (Section V-C).
  • K-means cluster count per corpus = 43k/29k/3k/29k/3.5k for English/French/Mandarin/German/Wolof; 14k for LibriSpeech dev-clean
    Taken from prior ES-KMeans and DPDP work; set to roughly 10 percent of the number of SylSeg-detected tokens (Section V-C).
  • Minimum segment duration = 5 frames
    Inherited from original ES-KMeans to avoid trivial short segments (Section V-C).
  • Maximum candidate boundaries per segment = 4
    Inherited from original ES-KMeans to bound the DP search (Section V-C).
  • Number of ES-KMeans+ iterations = 5
    Number of re-segmentation and clustering passes for ES-KMeans+ (Section V-C).
  • HuBERT layer for boundary detection = layer 9
    Based on Pasad et al. (2024) findings (Section V-C).
  • HuBERT layer for clustering embeddings = layer 12
    Based on development experiments (Section V-C).
assumptions (6)
  • domain assumption HuBERT frame features encode word-boundary information via cosine-dissimilarity peaks.
    Both methods rely on prominence-based boundaries from HuBERT features (Section III, following Pasad et al. 2024).
  • domain assumption Averaging and normalizing PCA-reduced HuBERT features inside a segment yields embeddings that preserve word identity well enough for clustering.
    Lexicon construction uses averaged, L2-normalized embeddings with K-means (Section III, Fig. 2c).
  • domain assumption K-means is an adequate clustering model for acoustic word embeddings.
    Both systems use FAISS K-means, and the claim that clustering is the limiting factor presumes this model family is representative (Sections III and VII).
  • domain assumption ZeroSpeech Track 2 metrics (R-value, token F1, NED, bitrate) and forced-alignment phonemic transcriptions measure word discovery quality validly.
    All quantitative conclusions are based on these benchmark metrics (Section V-B).
  • domain assumption The rule-based syllabifier produces adequate ground-truth syllable boundaries for the word-versus-syllable analysis.
    Section VI-C uses the syllabify tool on LibriSpeech phonemic transcriptions and notes its imperfections.
  • standard math The Viterbi dynamic program and the length-weighted scoring function d(z) correctly optimize the ES-KMeans objective.
    Equations (1) to (3) in Section IV-A define the objective and DP recursion; the length factor prevents trivial short segments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Should Top-Down Clustering Affect Boundaries in Unsupervised Word Discovery?." pith.science (2026). https://pith.science/paper/YEZANSJB

@misc{pith2026250719204,
  author       = {Pith},
  title        = {Pith review of: Should Top-Down Clustering Affect Boundaries in Unsupervised Word Discovery?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YEZANSJB}},
  note         = {Machine review of arXiv:2507.19204}
}
read the original abstract

We investigate the problem of segmenting unlabeled speech into word-like units and clustering these to create a lexicon. Prior work can be categorized into two frameworks. Bottom-up methods first determine boundaries and then cluster the fixed segmented words into a lexicon. In contrast, top-down methods incorporate information from the clustered words to inform boundary selection. However, it is unclear whether top-down information is necessary to improve segmentation. To explore this, we look at two similar approaches that differ in whether top-down clustering informs boundary selection. Our simple bottom-up strategy predicts word boundaries using the dissimilarity between adjacent self-supervised features, then clusters the resulting segments to construct a lexicon. Our top-down system is an updated version of the ES-KMeans dynamic programming method that iteratively uses K-means to update its boundaries. On the five-language ZeroSpeech benchmarks, both approaches achieve comparable state-of-the-art results, with the bottom-up system being nearly five times faster. Through detailed analyses, we show that the top-down influence of ES-KMeans can be beneficial (depending on factors like the candidate boundaries), but in many cases the simple bottom-up method performs just as well. For both methods, we show that the clustering step is a limiting factor. Therefore, we recommend that future work focus on improved clustering techniques and learning more discriminative word-like representations. Project code repository: https://github.com/s-malan/prom-seg-clus.

Figures

Figures reproduced from arXiv: 2507.19204 by the authors.

Figure 1
Figure 1. An example of word boundaries from the prominence-based approach [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. The dynamic programming segmentation step of ES-KMeans. Candidate [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Transcription of word units present in the largest clusters of (a) ES [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Lexicon quality of ES-KMeans+ (blue dot) and prominence segmenta [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

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. ZeroSyl: Simple Zero-Resource Syllable Tokenization for Spoken Language Modeling

    cs.CL 2026-02 conditional novelty 6.0 of 10

    Syllable boundaries can be derived directly from L2 norms of frozen WavLM features, yielding competitive spoken-language-model tokens without any training.

Reference graph

Works this paper leans on

38 extracted references · 36 canonical work pages · cited by 1 Pith paper

  1. [1]

    Computational modeling of phonetic and lexical learning in early language acquisition: Existing models and future directions,

    O. J. R ¨as¨anen, “Computational modeling of phonetic and lexical learning in early language acquisition: Existing models and future directions,” Speech Communication , vol. 54, pp. 975–997, 2012

    work page 2012

  2. [2]

    At 6–9 months, human infants know the meanings of many common nouns,

    E. Bergelson and D. Swingley, “At 6–9 months, human infants know the meanings of many common nouns,” National Academy of Sciences of the United States of America , vol. 190, pp. 3253–3258, 2012

    work page 2012

  3. [3]

    Some critical developments in acquiring native language sound organization during the first year,

    P. W. Jusczyk, “Some critical developments in acquiring native language sound organization during the first year,” Annals of Otology, Rhinology & Laryngology , vol. 189, pp. 11–15, 2002

    work page 2002

  4. [4]

    Cognitive science in the era of artificial intelligence: A roadmap for reverse-engineering the infant language-learner,

    E. Dupoux, “Cognitive science in the era of artificial intelligence: A roadmap for reverse-engineering the infant language-learner,” Cognition, vol. 173, pp. 43–59, 2016

    work page 2016

  5. [5]

    Automatic speech recognition for under-resourced languages: A survey,

    L. Besacier, E. Barnard, A. Karpov, and T. Schultz, “Automatic speech recognition for under-resourced languages: A survey,” Speech Communication, vol. 56, pp. 85–100, 2014

    work page 2014

  6. [6]

    Unsupervised pattern discovery in speech,

    A. S. Park and J. R. Glass, “Unsupervised pattern discovery in speech,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 16, pp. 186–197, 2008

    work page 2008

  7. [7]

    Unsupervised discovery of recurring speech patterns using probabilistic adaptive metrics,

    O. R¨as¨anen and M. A. C. Bland ´on, “Unsupervised discovery of recurring speech patterns using probabilistic adaptive metrics,” in Interspeech, 2020

    work page 2020

  8. [8]

    Spoken- term discovery using discrete speech units,

    B. van Niekerk, J. Za ¨ıdi, M.-A. Carbonneau, and H. Kamper, “Spoken- term discovery using discrete speech units,” in Interspeech, 2024. 10

    work page 2024

Show all 38 references
  1. [9]

    Unsupervised lexicon discovery from acoustic input,

    C.-y. Lee, T. O’Donnell, and J. R. Glass, “Unsupervised lexicon discovery from acoustic input,” Transactions of the Association for Computational Linguistics, vol. 3, pp. 389–403, 2015

    2015

  2. [10]

    DP-Parse: Finding word boundaries from raw speech with an instance lexicon,

    R. Algayres, T. Ricoul, J. Karadayi, H. Laurenc ¸on, S. Zaiem, A. Mohame, B. Sagot, and E. Dupoux, “DP-Parse: Finding word boundaries from raw speech with an instance lexicon,” Transactions of the Association for Computational Linguistics , vol. 10, pp. 1051–1065, 2022

    2022

  3. [11]

    Double articulation analyzer with prosody for unsupervised word and phone discovery,

    Y . Okuda, R. Ozaki, S. Komura, and T. Taniguchi, “Double articulation analyzer with prosody for unsupervised word and phone discovery,” IEEE Transactions on Cognitive and Developmental Systems , vol. 15, pp. 1335–1347, 2022

    2022

  4. [12]

    Unsupervised word discovery from speech using automatic segmentation into syllable-like units,

    O. J. R¨as¨anen, G. Doyle, and M. C. Frank, “Unsupervised word discovery from speech using automatic segmentation into syllable-like units,” in Interspeech, 2015

    2015

  5. [13]

    Self-expressing autoencoders for unsupervised spoken term discovery,

    S. Bhati, J. Villalba, P. ˙Zelasko, and N. Dehak, “Self-expressing autoencoders for unsupervised spoken term discovery,” in Interspeech, 2020

    2020

  6. [14]

    Word segmentation on discovered phone units with dynamic programming and self-supervised scoring,

    H. Kamper, “Word segmentation on discovered phone units with dynamic programming and self-supervised scoring,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 31, pp. 684–694, 2023

    2023

  7. [15]

    Revisiting speech segmentation and lexicon learning with better features,

    H. Kamper and B. van Niekerk, “Revisiting speech segmentation and lexicon learning with better features,” arXiv preprint arXiv:2401.17902 , 2024

    2024 arXiv

  8. [16]

    A segmental framework for fully-unsupervised large-vocabulary speech recognition,

    H. Kamper, A. Jansen, and S. Goldwater, “A segmental framework for fully-unsupervised large-vocabulary speech recognition,” Computer Speech & Language , vol. 46, pp. 154–174, 2017

    2017

  9. [17]

    Seg- mental contrastive predictive coding for unsupervised word segmentation,

    S. Bhati, J. Villalba, P. ˙Zelasko, L. Moro-Velazquez, and N. Dehak, “Seg- mental contrastive predictive coding for unsupervised word segmentation,” in Interspeech, 2021

    2021

  10. [18]

    Contrastive prediction strategies for unsupervised segmentation and categorization of phonemes and words,

    S. Cuervo, M. Grabias, J. Chorowski, G. Ciesielski, A. Ła´ncucki, P. Rychlikowski, and R. Marxer, “Contrastive prediction strategies for unsupervised segmentation and categorization of phonemes and words,” in ICASSP, 2022

    2022

  11. [19]

    An embedded segmental K-means model for unsupervised segmentation and clustering of speech,

    H. Kamper, K. Livescu, and S. Goldwater, “An embedded segmental K-means model for unsupervised segmentation and clustering of speech,” in ASRU, 2017

    2017

  12. [20]

    What do self- supervised speech models know about words?

    A. Pasad, C.-M. Chien, S. Settle, and K. Livescu, “What do self- supervised speech models know about words?” Transactions of the Association for Computational Linguistics , vol. 12, pp. 372–391, 2024

    2024

  13. [21]

    The Zero Resource Speech Challenge 2020: Discovering discrete subword and word units,

    E. Dunbar, J. Karadayi, M. Bernard, X.-N. Cao, R. Algayres, L. Ondel, L. Besacier, S. Sakti, and E. Dupoux, “The Zero Resource Speech Challenge 2020: Discovering discrete subword and word units,” in Interspeech, 2020

    2020

  14. [22]

    Unsupervised word discovery: Boundary detection with clustering vs. dynamic programming,

    S. Malan, B. van Niekerk, and H. Kamper, “Unsupervised word discovery: Boundary detection with clustering vs. dynamic programming,” in ICASSP, 2025

    2025

  15. [23]

    Self-supervised speech representation learning: A review,

    A. Mohamed, H.-y. Lee, L. Borgholt, J. D. Havtorn, J. Edin, C. Igel, K. Kirchhoff, S.-W. Li, K. Livescu, L. Maaløe, T. N. Sainath, and S. Watanabe, “Self-supervised speech representation learning: A review,” IEEE Journal of Selected Topics in Signal Processing , vol. 16, pp. 1...

    2022

  16. [25]

    A computational model for unsupervised word discovery,

    L. ten Bosch and B. Cranen, “A computational model for unsupervised word discovery,” in InterSpeech, 2007

    2007

  17. [26]

    Unsupervised word discovery from speech using automatic segmentation into syllable-like units,

    O. J. R ¨as¨anen, G. Doyle, and M. Frank, “Unsupervised word discovery from speech using automatic segmentation into syllable-like units,” in Interspeech, 2015

    2015

  18. [27]

    Gibbs sampling for the uninitiated,

    P. Resnik and E. Hardisty, “Gibbs sampling for the uninitiated,” Tech. Rep., University of Maryland, College Park, MD , 2010

    2010

  19. [28]

    HuBERT: Self-supervised speech representation learning by masked prediction of hidden units,

    W. Hsu, B. Bolte, Y . H. Tsai, K. Lakhotia, R. Salakhutdinov, and A. Mohamed, “HuBERT: Self-supervised speech representation learning by masked prediction of hidden units,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 29, pp. 3451–3460, 2021

    2021

  20. [30]

    Analyzing acoustic word embeddings from pre-trained self-supervised speech models,

    R. Sanabria, H. Tang, and S. Goldwater, “Analyzing acoustic word embeddings from pre-trained self-supervised speech models,” in ICASSP, 2023

    2023

  21. [31]

    Word- level acoustic modeling with convolutional vector regression,

    A. L. Maas, S. D. Miller, T. M. O’neil, A. Y . Ng, and P. Nguyen, “Word- level acoustic modeling with convolutional vector regression,” in ICML, 2012

    2012

  22. [32]

    Fixed-dimensional acoustic embeddings of variable-length segments in low-resource settings,

    K. Levin, K. Henry, A. Jansen, and K. Livescu, “Fixed-dimensional acoustic embeddings of variable-length segments in low-resource settings,” in ASRU, 2013

    2013

  23. [33]

    Unsupervised lexical clustering of speech segments using fixed-dimensional acoustic embeddings,

    H. Kamper, A. Jansen, S. King, and S. Goldwater, “Unsupervised lexical clustering of speech segments using fixed-dimensional acoustic embeddings,” in SLT, 2014

    2014

  24. [34]

    Word discovery in visually grounded, self- supervised speech models,

    P. Peng and D. Harwath, “Word discovery in visually grounded, self- supervised speech models,” in Interspeech, 2022

    2022

  25. [35]

    LibriSpeech: An ASR corpus based on public domain audio books,

    V . Panayotov, G. Chen, D. Povey, and S. Khudanpur, “LibriSpeech: An ASR corpus based on public domain audio books,” in ICASSP, 2015

    2015

  26. [36]

    An improved speech segmentation quality measure: the R-value,

    O. J. R ¨as¨anen, U. K. Laine, and T. Altosaar, “An improved speech segmentation quality measure: the R-value,” in Interspeech, 2009

    2009

  27. [37]

    On the robust automatic segmentation of spontaneous speech,

    B. Petek, O. Andersen, and P. Dalsgaard, “On the robust automatic segmentation of spontaneous speech,” in ICSLP, 1996

    1996

  28. [38]

    Bridging the gap between speech technology and natural language processing: an evaluation toolbox for term discovery systems,

    B. Ludusan, M. Versteegh, A. Jansen, G. Gravier, X.-N. Cao, M. Johnson, and E. Dupoux, “Bridging the gap between speech technology and natural language processing: an evaluation toolbox for term discovery systems,” in LREC, 2014

    2014

  29. [39]

    A comparison of discrete and soft speech units for improved voice conversion,

    B. van Niekerk, M.-A. Carbonneau, J. Zaidi, M. Baas, H. Seute, and H. Kamper, “A comparison of discrete and soft speech units for improved voice conversion,” in ICASSP, 2022

    2022

  30. [40]

    The Zero Resource Speech Challenge 2019: TTS without T,

    E. Dunbar, R. Algayres, J. Karadayi, M. Bernard, J. Benjumea, X.-N. Cao Kam, L. Miskic, C. Dugrain, L. Ondel, A. Black, L. Besacier, S. Sakti, and E. Dupoux, “The Zero Resource Speech Challenge 2019: TTS without T,” in Interspeech, 2019

    2019

Pith tools

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