Pith. sign in

REVIEW 4 major objections 5 minor 36 references

MSA-ASR: Efficient Multilingual Speaker Attribution with frozen ASR Models

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Freezing a multilingual ASR while training only a speaker module yields speaker-attributed transcripts across languages without ASR fine-tuning.

desk verdict A useful frozen-ASR recipe for multilingual speaker attribution that works for non-overlapping and lightly overlapping speech, but the abstract's overlap claim is contradicted by the paper's own Table III. read the letter →

arxiv 2411.18152 v2 pith:GR3EEOTE submitted 2024-11-27 cs.CL cs.SDeess.AS

Thai-Binh Nguyen , Alexander Waibel This is my paper · ORCID
classification cs.CLcs.SDeess.AS
keywords speaker-attributedautomaticspeechrecognitionmultilingualASRfrozenmodelsspeakerembeddingsweaklabelsembeddingalignmentanddiscriminationlosscpWERspectralclustering
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that speaker-attributed ASR does not require modifying the ASR model. It proposes to freeze a multilingual ASR and train a separate speaker module that predicts a speaker embedding for every recognized token, using weak labels built from a pre-trained speaker-embedding model on ordinary monolingual ASR data. If the claim holds, any robust frozen ASR can be upgraded with speaker attribution while keeping its recognition behavior unchanged, and the attribution transfers to languages and recording conditions the speaker module never saw during training. That matters because existing joint SA-ASR systems require fine-tuning the ASR on language-specific multi-speaker data, which is scarce.

What carries the argument

The load-bearing components are: a frozen transformer ASR whose encoder features are reused by the speaker decoder via cross-attention; a speaker decoder that predicts one speaker embedding per ASR token, with word and position embeddings shared with the ASR decoder so outputs align; an Embedding Alignment and Discrimination (EAD) loss, Eq. (7), which is a weighted sum of a cosine alignment loss between predicted and target embeddings, an MSE loss making predicted pairwise similarities match target pairwise similarities, and an MSE loss making predicted-to-target similarities match target pairwise similarities; and weak labels built by embedding each speaker turn with a pre-trained speaker-embedding model and grouping turns by cosine similarity above a threshold of 0.7. Spectral clustering on the predicted token embeddings produces the final speaker assignment. The EAD loss is what lets the model learn to separate speakers without ground-truth speaker labels.

What would settle it

Shuffle the target speaker embeddings in the training targets so every transcript is paired with the embedding of a different speaker, retrain the speaker module, and compare cpWER on the 16-language benchmark; if the score stays about the same, the attribution is not coming from the weak speaker labels. Alternatively, build same-speaker and different-speaker cosine-similarity histograms with the pre-trained speaker-embedding model on the target languages: heavy overlap near the 0.7 threshold would show the labeling step cannot separate speakers there.

Watch

Extended reading notes

Core claim

The central claim is that training only the speaker module on non-overlapping, monolingual, synthetic multi-turn data produces a model that assigns transcribed words to speakers across languages, including on data with overlapping speech, without any fine-tuning of the ASR component. The speaker module is trained to match speaker-embedding targets computed for each speaker turn, then at inference its token-level embeddings are clustered with spectral clustering to assign speakers. On a 16-language benchmark, the system raises cpWER by only 29% relative over ASR alone, versus 92% for a diarization-plus-ASR baseline; on AMI-IHM it is competitive with fine-tuned joint baselines, and on a mixed-language meeting set it roughly halves cpWER relative to diarization-plus-ASR. The paper also shows the speaker module can run independently of the ASR and can consume gold transcripts, producing very low cpWER when transcripts are provided.

Load-bearing premise

The method assumes that the pre-trained speaker-embedding model separates speakers well enough in every target language and domain that grouping turns by a fixed cosine threshold of 0.7 yields correct weak labels; if that fails, the training targets are wrong and the whole attribution collapses.

Editorial extensions

If this is right

  • A frozen ASR can be turned into an SA-ASR system by training only a speaker module, preserving the ASR's original recognition performance and multilingual coverage.
  • Attribution trained on English non-overlapping data transfers to 16 languages and to a four-language mixed meeting corpus with no in-domain fine-tuning.
  • On non-overlapping and low-overlap conditions, the approach beats a modular diarization-plus-ASR baseline and matches fine-tuned joint baselines on AMI-IHM; on high-overlap conditions it degrades because the ASR is frozen.
  • The speaker module can be run independently of the ASR decoder and can even accept an external gold transcript, which yields near-perfect speaker attribution when a correct transcript is available.

Reading between the lines

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

  • Editorial inference: the same weak-label recipe should work with other frozen ASR models, since the speaker decoder only needs access to encoder features and shared token embeddings; the paper demonstrates it with one specific ASR.
  • Editorial inference: the cosine threshold is a hyperparameter fixed at 0.7 without target-language calibration, so adapting it from the similarity distribution of the pre-trained speaker-embedding model is a natural extension.
  • Editorial inference: because the speaker decoder is independent of the ASR decoder, the design is a candidate for streaming or low-latency attribution where a parallel speaker head runs on the ASR encoder output; the paper does not explore this.
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 proposes MSA-ASR, a speaker-attributed automatic speech recognition system that keeps a frozen Whisper large-v2 ASR model and trains only a speaker module (encoder/decoder) to predict token-level speaker embeddings. Training is performed on concatenated non-overlapping monolingual turns using weak speaker labels produced by TitaNet-L; an Embedding Alignment and Discrimination (EAD) loss aligns the predicted embedding sequence with the weak-label sequence and preserves within-sequence pairwise similarity structure. At inference, spectral clustering assigns speakers. The system is evaluated with cpWER on VoxPopuli (16 languages, non-overlapping multi-talker), AMI-IHM, LibriCSS, and a 45-minute mixed-language dataset. The central result is that on VoxPopuli, MSA-ASR incurs only a 29.3% relative cpWER increase over plain ASR, versus 92% for the Pyannote-diarization-plus-ASR baseline, and it is competitive on AMI with gold VAD. The paper also reports releasing the model and dataset.

Significance. If the claims hold, the paper offers a practical recipe for multilingual speaker-attributed ASR: keep a frozen, broadly trained ASR model and add a small trainable speaker module, avoiding per-language fine-tuning of the ASR. The VoxPopuli results are strong and show a clear advantage over a standard diarization-plus-ASR baseline across languages. The AMI result is competitive under gold VAD, and the idea of using weak speaker embeddings as training targets is interesting. However, the headline claim about overlapping speech is not supported by the reported LibriCSS numbers, and the only multilingual overlap evaluation is tiny and nearly overlap-free. The release of the pre-trained model and dataset is a concrete contribution that supports reproducibility. Overall the core method appears promising for non-overlapping and lightly overlapping speech, but the manuscript overstates its overlap robustness.

major comments (4)
  1. [Abstract and Section III-C, Table III] The abstract claims that the approach 'effectively extracts speaker attributes across diverse multilingual datasets, including those with overlapping speech,' but Table III shows the opposite for all non-trivial overlap conditions: MSA-ASR cpWER is 27.9 at 20% overlap, 41.7 at 30%, and 46.5 at 40%, while the NME-SC + SOT-SA-ASR baseline stays at 13.9 even at 40% overlap. The text itself states that the degradation is 'expected' because the ASR is frozen. This is a direct mismatch between the headline claim and the reported evidence. The manuscript must either provide overlap-robust results (e.g., by adding an overlap-handling mechanism or evaluating on a stronger overlap-aware benchmark) or explicitly re-scope the claim to non-overlapping and lightly overlapping speech.
  2. [Section III-A, mixed-language dataset] The only multilingual evaluation involving overlapping speech is the in-house mixed-language dataset, which totals 45 minutes and has an overlap rate of approximately 3%. This is far too small and too low-overlap to support a general statement about multilingual overlapping-speech robustness. At minimum, the authors should report the amount of overlapping audio in each condition, provide overlap-stratified results, or acknowledge that the multilingual overlap claim is not established by this evaluation.
  3. [Section II-B] The training targets T are produced by TitaNet-L embeddings, and similar turns are grouped by a cosine threshold theta=0.7, but the paper gives no validation of this weak-label premise: no analysis of TitaNet-L embedding quality on the training languages/domains, no threshold sensitivity study, and no failure analysis. Since the EAD loss in Eq. (7) is trained entirely against these targets, miscalibrated embeddings or an inappropriate threshold would directly degrade speaker attribution. The authors should report at least a sensitivity analysis for theta and a sanity check of the target embeddings (e.g., same-speaker vs. different-speaker similarity distributions).
  4. [Section II-B and III-B] The training corpus is never named: the paper says the model is trained 'only on the data that has been processed as described in section II-B' using 'standard monolingual ASR datasets,' but it does not specify which corpus or corpora, their language composition, or the train/dev split. Because the central selling point is zero fine-tuning on target languages, reproducibility requires naming the exact training data and its overlap/language statistics.
minor comments (5)
  1. [Throughout] There are several typographical inconsistencies, including 'V oxpopuli' and 'V AD' (with a stray space) in the text and tables, and the reference [29] title spells 'Titanet' while the body uses 'TitaNet-L.' These should be corrected.
  2. [Table III] The column headers '0S 0L' are not defined in the caption or text; please explain what S and L denote so the reader can interpret the zero-overlap baseline conditions.
  3. [Section III-A] For the VoxPopuli multi-talker setup, the description says samples have 'an average of 2.5 speakers' and 'up to 5 non-overlapping turns,' but it is unclear whether speakers can repeat across turns and how the 2.5 average is computed. Please clarify the data construction to allow replication.
  4. [Section III-B and Eq. (7)] The loss weights are set to alpha = beta = gamma = 1 with no discussion or ablation. Given that the EAD loss is central to training, a small ablation or at least a statement that the weights were not tuned would strengthen the paper.
  5. [Section III-C, Table IV] The two-row gain of MSA-ASR over NME-SC + SOT-SA-ASR (15.6 vs. 16.3 dev, 14.3 vs. 15.1 eval) is presented as competitive, which is fair, but without error bars or significance tests the reader cannot tell whether the small differences are meaningful. Please add variance estimates or multiple runs.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: MSA-ASR is benchmarked externally, and the TitaNet-L-derived training targets and the self-citation of [23] are training and citation choices, not circular reductions.

full rationale

The central claim of MSA-ASR is that a frozen Whisper model plus a separately trained speaker module can perform speaker-attributed ASR in multiple languages despite being trained only on non-overlapping monolingual data. This claim is evaluated against external benchmarks (cpWER on VoxPopuli, LibriCSS, AMI-IHM, and an in-house mixed-language set), not against the training targets themselves. The speaker module is trained with the EAD loss (Eqs. 1-7) to regress TitaNet-L embeddings, but the reported cpWER numbers are computed from ASR transcripts and reference speaker labels using spectral clustering; no reported number is obtained by re-inserting a fitted constant into the training objective. Using TitaNet-L both to build the weak-label targets and to group similar turns (Section II-B) is a training-signal choice, not a circular derivation: the evaluation would be circular only if the metric were similarity to TitaNet-L outputs, which it is not. The self-citation of [23] supplies the turn-concatenation recipe and noise/reverberation augmentation, but the distinct contribution of this paper is token-level speaker embedding prediction with the EAD loss under a frozen multilingual ASR, and that contribution is independently evaluated. The abstract's claim about overlapping speech is stronger than the evidence: Table III shows cpWER rising from 11.5 at 10% overlap to 46.5 at 40%, and the text admits this drop is 'expected' because the ASR is frozen. That is a claim-data mismatch and a correctness risk, not a circular step. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known result is repackaged. The score of 1 reflects only the mild self-referential use of TitaNet-L for both targets and grouping; no actual circular step is present.

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

The central claim rests on external pre-trained models (Whisper large-v2 for ASR, TitaNet-L for speaker embeddings), synthetic non-overlapping training data, and hand-set hyperparameters. No new physical, linguistic, or architectural entity is introduced beyond the EAD loss, which is a training objective rather than an invented entity.

free parameters (3)
  • theta (similarity threshold) = 0.7
    Section II-B: speaker turns are grouped as similar when TitaNet-L cosine similarity exceeds theta. The value is chosen by hand and no ablation is reported.
  • EAD loss weights alpha, beta, gamma = 1, 1, 1
    Section III-B: all three loss terms in Eq. (7) are weighted equally. No tuning study is presented.
  • K (number of speaker decoder layers using Hasr as key) = 1
    Sections II-A and III-B: the first K=1 layers of the Speaker decoder use ASR hidden features as key, the remaining layers use speaker hidden features. This architectural hyperparameter is set by hand.
assumptions (5)
  • domain assumption Whisper large-v2 is a sufficient frozen multilingual ASR backbone for the speaker-attributed task.
    Section II-A and III-B use Whisper large-v2 without modification, and the paper's claims depend on its multilingual recognition quality.
  • domain assumption TitaNet-L embeddings reliably identify the same speaker across turns and languages.
    Section II-B uses TitaNet-L to compute target speaker embeddings and to choose similar turns; if this model is miscalibrated in target domains, the weak labels are wrong.
  • domain assumption Random concatenation of non-overlapping English turns, plus noise and reverberation, approximates real multi-talker distributions including overlap and multilingual speech.
    Section II-B constructs training samples from non-overlapping turns, yet Section III-C claims generalization to overlapping and multilingual audio.
  • domain assumption Sharing word and position embeddings between the ASR and Speaker decoders ensures token-to-speaker alignment.
    Section II-A states that embeddings are shared 'to ensure alignment between their outputs'; the model provides no other alignment mechanism.
  • domain assumption Spectral clustering on pairwise cosine similarities of predicted embeddings yields correct speaker assignments.
    Section III-B uses spectral clustering for speaker assignment; the paper does not validate clustering accuracy separately from cpWER.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MSA-ASR: Efficient Multilingual Speaker Attribution with frozen ASR Models." pith.science (2026). https://pith.science/paper/GR3EEOTE

@misc{pith2026241118152,
  author       = {Pith},
  title        = {Pith review of: MSA-ASR: Efficient Multilingual Speaker Attribution with frozen ASR Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GR3EEOTE}},
  note         = {Machine review of arXiv:2411.18152}
}
read the original abstract

Speaker-attributed automatic speech recognition (SA-ASR) aims to transcribe speech while assigning transcripts to the corresponding speakers accurately. Existing methods often rely on complex modular systems or require extensive fine-tuning of joint modules, limiting their adaptability and general efficiency. This paper introduces a novel approach, leveraging a frozen multilingual ASR model to incorporate speaker attribution into the transcriptions, using only standard monolingual ASR datasets. Our method involves training a speaker module to predict speaker embeddings based on weak labels without requiring additional ASR model modifications. Despite being trained exclusively with non-overlapping monolingual data, our approach effectively extracts speaker attributes across diverse multilingual datasets, including those with overlapping speech. Experimental results demonstrate competitive performance compared to strong baselines, highlighting the model's robustness and potential for practical applications.

Figures

Figures reproduced from arXiv: 2411.18152 by the authors.

Figure 1
Figure 1. Overview of MSA-ASR model. ASR decoder processes tokens [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Speaker decoder architecture. Similar as standard transformer decoder, [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 33 canonical work pages

  1. [1]

    Mul- timodal people id for a multimedia meeting browser,

    J. Yang, X. Zhu, R. Gross, J. Kominek, Y . Pan, and A. Waibel, “Mul- timodal people id for a multimedia meeting browser,” in Proceedings of the seventh ACM international conference on Multimedia (Part 1) , 1999, pp. 159–168

    work page 1999

  2. [2]

    Towards a multimodal meeting record,

    R. Gross, M. Bett, H. Yu, X. Zhu, Y . Pan, J. Yang, and A. Waibel, “Towards a multimodal meeting record,” in 2000 IEEE International Conference on Multimedia and Expo. ICME2000. Proceedings. Lat- est Advances in the Fast Changing World of Multimedia (Cat. No. 00TH8532), vol. 3. IEEE, 2000, pp. 1593–1596

    work page 2000

  3. [3]

    Estimating focus of attention based on gaze and sound,

    R. Stiefelhagen, J. Yang, and A. Waibel, “Estimating focus of attention based on gaze and sound,” in Proceedings of the 2001 workshop on Perceptive user interfaces , 2001, pp. 1–9

    work page 2001

  4. [4]

    Estimating focus of attention based on gaze and sound,

    R. S, J. Yang, and A. Waibel, “Estimating focus of attention based on gaze and sound,” in Workshop on Perceptive User Interfaces . Associ- ation for Computing Machinery, 2001

    work page 2001

  5. [5]

    Chil: Computers in the human interaction loop,

    A. Waibel, H. Steusloff, R. Stiefelhagen et al. , “Chil: Computers in the human interaction loop,” 2005

    work page 2005

  6. [6]

    Simultaneous translation of open domain lectures and speeches,

    A. Waibel and C. Fuegen, “Simultaneous translation of open domain lectures and speeches,” 2012, uS Patent 8,090,570

    work page 2012

  7. [7]

    Integration of speech separation, diarization, and recognition for multi-speaker meetings: System de- scription, comparison, and analysis,

    D. Raj, P. Denisov, Z. Chen, H. Erdogan, Z. Huang, M. He, S. Watanabe, J. Du, T. Yoshioka, Y . Luo et al. , “Integration of speech separation, diarization, and recognition for multi-speaker meetings: System de- scription, comparison, and analysis,” in 2021 IEEE spoken language technology workshop (SLT) . IEEE, 2021, pp. 897–904

    work page 2021

  8. [8]

    A comparative study of modular and joint approaches for speaker-attributed asr on monaural long-form audio,

    N. Kanda, X. Xiao, J. Wu, T. Zhou, Y . Gaur, X. Wang, Z. Meng, Z. Chen, and T. Yoshioka, “A comparative study of modular and joint approaches for speaker-attributed asr on monaural long-form audio,” in 2021 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) . IEEE, 2021, pp. 296–303

    work page 2021

Show all 36 references
  1. [9]

    Improving speaker assignment in speaker-attributed asr for real meeting applications,

    C. Cui, I. A. Sheikh, M. Sadeghi, and E. Vincent, “Improving speaker assignment in speaker-attributed asr for real meeting applications,” in The Speaker and Language Recognition Workshop , 2024

    2024

  2. [10]

    A Comparative Study on Speaker-attributed Automatic Speech Recognition in Multi-party Meetings,

    F. Yu, Z. Du, S. Zhang, Y . Lin, and L. Xie, “A Comparative Study on Speaker-attributed Automatic Speech Recognition in Multi-party Meetings,” in Proc. Interspeech 2022 , 2022

    2022

  3. [11]

    Convoifilter: A case study of doing cock- tail party speech recognition,

    T.-B. Nguyen and A. Waibel, “Convoifilter: A case study of doing cock- tail party speech recognition,” in 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW) , 2024, pp. 565–569

    2024

  4. [12]

    Investigation of end-to-end speaker-attributed asr for continuous multi-talker recordings,

    N. Kanda, X. Chang, Y . Gaur, X. Wang, Z. Meng, Z. Chen, and T. Yoshioka, “Investigation of end-to-end speaker-attributed asr for continuous multi-talker recordings,” in 2021 IEEE Spoken Language Technology Workshop (SLT). IEEE, 2021, pp. 809–816

    2021

  5. [13]

    Transcribe-to-diarize: Neural speaker diarization for unlimited number of speakers using end-to-end speaker-attributed asr,

    N. Kanda, X. Xiao, Y . Gaur, X. Wang, Z. Meng, Z. Chen, and T. Yosh- ioka, “Transcribe-to-diarize: Neural speaker diarization for unlimited number of speakers using end-to-end speaker-attributed asr,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and S...

    2022

  6. [14]

    Streaming speaker-attributed asr with token-level speaker embeddings,

    N. Kanda, J. Wu, Y . Wu, X. Xiao, Z. Meng, X. Wang, Y . Gaur, Z. Chen, J. Li, and T. Yoshioka, “Streaming speaker-attributed asr with token-level speaker embeddings,” in Interspeech, 2022

    2022

  7. [15]

    Sa- paraformer: Non-autoregressive end-to-end speaker-attributed asr,

    Y . Li, F. Yu, Y . Liang, P. Guo, M. Shi, Z. Du, S. Zhang, and L. Xie, “Sa- paraformer: Non-autoregressive end-to-end speaker-attributed asr,” in 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). IEEE, 2023, pp. 1–7

    2023

  8. [16]

    End-to-End Speaker-Attributed ASR with Transformer,

    N. Kanda, G. Ye, Y . Gaur, X. Wang, Z. Meng, Z. Chen, and T. Yosh- ioka, “End-to-End Speaker-Attributed ASR with Transformer,” in Proc. Interspeech 2021 , 2021, pp. 4413–4417

    2021

  9. [17]

    CASA-ASR: Context-Aware Speaker-Attributed ASR,

    M. Shi, Z. Du, Q. Chen, F. Yu, Y . Li, S. Zhang, J. Zhang, and L.-R. Dai, “CASA-ASR: Context-Aware Speaker-Attributed ASR,” in Proc. INTERSPEECH 2023 , 2023, pp. 411–415

    2023

  10. [18]

    One model to rule them all? towards end-to-end joint speaker diarization and speech recognition,

    S. Cornell, J.-w. Jung, S. Watanabe, and S. Squartini, “One model to rule them all? towards end-to-end joint speaker diarization and speech recognition,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 11 ...

    2024

  11. [19]

    Robust speech recognition via large-scale weak supervi- sion,

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

    2023

  12. [20]

    Speaker identification using mul- tilingual phone strings,

    Q. Jin, T. Schultz, and A. Waibel, “Speaker identification using mul- tilingual phone strings,” in 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing , vol. 1. IEEE, 2002, pp. I–145

    2002

  13. [21]

    Phonetic speaker identification

    ——, “Phonetic speaker identification.” in INTERSPEECH, 2002, pp. 1345–1348

    2002

  14. [22]

    How multilingual is multilingual BERT?

    T. Pires, E. Schlinger, and D. Garrette, “How multilingual is multilingual BERT?” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, A. Korhonen, D. Traum, and L. M`arquez, Eds. Florence, Italy: Association for Computational Linguistics...

    2019

  15. [23]

    Synthetic conversations improve multi- talker asr,

    T.-B. Nguyen and A. Waibel, “Synthetic conversations improve multi- talker asr,” in ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2024, pp. 10 461– 10 465

    2024

  16. [24]

    Simulating realistic speech overlaps improves multi- talker asr,

    M. Yang, N. Kanda, X. Wang, J. Wu, S. Sivasankaran, Z. Chen, J. Li, and T. Yoshioka, “Simulating realistic speech overlaps improves multi- talker asr,” in ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2023, pp. 1–5

    2023

  17. [25]

    Overlaps and gender analysis in the context of broadcast media,

    M. Lebourdais, M. Tahon, A. Laurent, S. Meignier, and A. Larcher, “Overlaps and gender analysis in the context of broadcast media,” in Proceedings of the Thirteenth Language Resources and Evaluation Conference, N. Calzolari, F. B ´echet, P. Blache, K. Choukri, C. Cieri, T. Dec...

    2022

  18. [26]

    Detecting double-talk (overlapping speech) in conversations using deep learning,

    Abdullah, “Detecting double-talk (overlapping speech) in conversations using deep learning,” 2017. [Online]. Available: https://publica. fraunhofer.de/handle/publica/281843

    2017

  19. [27]

    Detecting and counting overlapping speakers in distant speech scenarios,

    S. Cornell, M. Omologo, S. Squartini, and E. Vincent, “Detecting and counting overlapping speakers in distant speech scenarios,” in Interspeech, 2020, pp. 3107–3111

    2020

  20. [28]

    A comparison of metric learning loss functions for end-to-end speaker verification,

    J. M. Coria, H. Bredin, S. Ghannay, and S. Rosset, “A comparison of metric learning loss functions for end-to-end speaker verification,” in In- ternational Conference on Statistical Language and Speech Processing . Springer, 2020, pp. 137–148

    2020

  21. [29]

    Titanet: Neural model for speaker representation with 1d depth-wise separable convolutions and global context,

    N. R. Koluguri, T. Park, and B. Ginsburg, “Titanet: Neural model for speaker representation with 1d depth-wise separable convolutions and global context,” in ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2022, pp. 8102– 8106

    2022

  22. [30]

    V oxPopuli: A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation,

    C. Wang, M. Riviere, A. Lee, A. Wu, C. Talnikar, D. Haziza, M. Williamson, J. Pino, and E. Dupoux, “V oxPopuli: A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation,” in Proceedings of the 59th Annual Meeting of the ...

    2021

  23. [31]

    The ami meeting corpus: A pre-announcement,

    J. Carletta, S. Ashby, S. Bourban, M. Flynn, M. Guillemot, T. Hain, J. Kadlec, V . Karaiskos, W. Kraaij, M. Kronenthal et al. , “The ami meeting corpus: A pre-announcement,” in International workshop on machine learning for multimodal interaction . Springer, 2005, pp. 28– 39

    2005

  24. [32]

    Continuous speech separation: Dataset and analysis,

    Z. Chen, T. Yoshioka, L. Lu, T. Zhou, Z. Meng, Y . Luo, J. Wu, X. Xiao, and J. Li, “Continuous speech separation: Dataset and analysis,” in ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2020, pp. 7284–7288

    2020

  25. [33]

    Chime-6 challenge: Tackling multispeaker speech recognition for unsegmented recordings,

    S. Watanabe, M. Mandel, J. Barker, and E. Vincent, “Chime-6 challenge: Tackling multispeaker speech recognition for unsegmented recordings,” CHiME 2020 , 2020

    2020

  26. [34]

    pyannote.audio 2.1 speaker diarization pipeline: principle, benchmark, and recipe,

    H. Bredin, “pyannote.audio 2.1 speaker diarization pipeline: principle, benchmark, and recipe,” in Proc. INTERSPEECH 2023 , 2023

    2023

  27. [35]

    Speaker diarization with lstm,

    Q. Wang, C. Downey, L. Wan, P. A. Mansfield, and I. L. Moreno, “Speaker diarization with lstm,” in 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2018, pp. 5239– 5243

    2018

  28. [36]

    Silero vad: pre-trained enterprise-grade voice activity detector (vad), number detector and language classifier,

    S. Team, “Silero vad: pre-trained enterprise-grade voice activity detector (vad), number detector and language classifier,” 2021

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.