Pith. sign in

REVIEW 3 major objections 3 minor 34 references

Speaker-Distinguishable CTC: Learning Speaker Distinction Using CTC for Multi-Talker Speech Recognition

T0 review · 3 major / 3 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 This paper proposes Speaker-Distinguishable CTC, a CTC extension that predicts both a token and a speaker per frame, and shows that adding its loss to Serialized Output Training cuts the multi-talker word error rate on two-speaker…

desk verdict A clean, parameter-light extension of CTC that appears to improve speaker assignment in synthetic two-talker mixtures, but the one-frame-one-speaker assumption is never stress-tested on real overlap. read the letter →

arxiv 2506.07515 v1 pith:4JEBIBFJ submitted 2025-06-09 eess.AS cs.CLcs.SD

Asahi Sakuma , Hiroaki Sato , Ryuga Sugano , Tadashi Kumano , Yoshihiko Kawai , Tetsuji Ogawa This is my paper
classification eess.AScs.CLcs.SD
keywords multi-talkerspeechrecognitionconnectionisttemporalclassificationserializedoutputtrainingspeakerattributionoverlappingend-to-endASRLibriMixspeaker-distinguishableCTC
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 sets out to show that the main failure in Serialized Output Training (SOT) for overlapping speech is speaker misassignment, not token misalignment, and that this failure can be repaired with a small CTC-style auxiliary loss using only the same data SOT already uses: overlapping audio plus each speaker's transcript. It introduces Speaker-Distinguishable CTC (SD-CTC), which extends each frame's token posterior with a speaker posterior, so every frame is assigned a token and a speaker, plus a speaker-specific blank that lets non-target speech be skipped. Trained jointly with SOT, SD-CTC reduces the concatenated minimum-permutation word error rate (cpWER) on the two-speaker LibriSpeechMix test set from 4.7% to 3.5%, a 26% relative reduction, and matches a state-of-the-art method that needs token-level timestamps while using fewer parameters. If true, this matters because real conversational corpora rarely come with timestamps, so multi-talker ASR could improve without auxiliary supervision.

What carries the argument

The load-bearing object is the SD-CTC loss, a per-speaker CTC loss computed from a joint token-speaker frame posterior. For speaker $\sigma$ and token $\rho$, the posterior is $P(\sigma,\rho\mid x_t)=P_s(\sigma\mid x_t)P_v(\rho\mid x_t)$ when $\rho$ is a real token, and $P(\sigma,\langle\neg s\rangle\mid x_t)=P_s(\sigma\mid x_t)P_v(\langle b\rangle\mid x_t)+(1-P_s(\sigma\mid x_t))$ for the speaker-specific blank; summing $-\log P(\sigma,\mathbf{y}_\sigma\mid X)$ over speakers gives $\mathcal{L}_{\mathrm{SD\text{-}CTC}}$. A two-stage schedule first pre-trains the CTC/attention model on single-speaker data with the speaker layer frozen at $P_s(s_1\mid x_t)=1$, then fine-tunes on multi-talker data with the token layer frozen, so the frozen token posterior acts as a per-frame weight for speaker training. This forces the encoder to separate speakers without any auxiliary information beyond overlapping speech and transcripts.

What would settle it

Evaluate SOT+SD-CTC against the SOT baseline on a real overlapping-conversation corpus where both speakers are genuinely audible in the same time frames rather than on synthetic additive mixtures; if the 26% relative cpWER reduction shrinks or disappears as true simultaneous speech dominates, the one-speaker-per-frame assumption is the breaking point.

Watch

Extended reading notes

Core claim

The central claim is that SOT models already place tokens at the right acoustic frames but frequently assign those tokens to the wrong speaker, and SD-CTC targets that failure directly. The model factorizes the per-frame distribution into a token posterior $P_v(\pi\mid x_t)$ and a speaker posterior $P_s(\sigma\mid x_t)$, defines the probability that speaker $\sigma$ emits token $\rho$ as $P_s(\sigma\mid x_t)P_v(\rho\mid x_t)$, and absorbs non-target frames into a speaker-specific blank $\langle\neg s\rangle$ with probability $P_s(\sigma\mid x_t)P_v(\langle b\rangle\mid x_t)+(1-P_s(\sigma\mid x_t))$. Summing the resulting CTC loss over speakers and using it as a multi-task loss with SOT makes the encoder separate speakers at the frame level; the authors report 3.5% cpWER on LibriSpeechMix versus 4.7% for SOT alone, comparable to 3.4% for SA-SOT with fewer parameters.

Load-bearing premise

The method assumes every audio frame belongs to exactly one speaker, even when two people are actually talking at once, so the per-frame speaker prediction always picks a single owner.

Editorial extensions

If this is right

  • On LibriSpeechMix, SOT plus SD-CTC cuts cpWER from 4.7% to 3.5%, a 26% relative reduction, using only overlapping speech and per-speaker transcripts.
  • The 3.5% result is comparable to SA-SOT's 3.4%, which requires token-level timestamps, and uses 114M parameters versus 136M.
  • It outperforms the other auxiliary-free CTC variants compared in the paper: SACTC (5.4%) and GEncSep (3.9%), with fewer parameters than GEncSep.
  • Because SD-CTC reduces to standard CTC when a single speaker is present, the added loss does not disturb single-speaker performance (2.4% on LibriSpeech test-clean).
  • SD-CTC adds only a linear speaker-prediction layer, so extending to more speakers changes just the output dimension of that layer, not the model architecture.

Reading between the lines

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

  • The paper leaves implicit that SD-CTC's one-speaker-per-frame factorization is most defensible for lightly overlapped speech; in dense real overlaps where both speakers are audible in the same frames, the per-frame speaker posterior cannot represent both, so the method would likely need a multi-label or mixture posterior.
  • Because the token posterior $P_v$ is frozen from single-speaker pre-training, a natural testable extension is to unfreeze or adapt it on mixed audio; this would show how much of the gain depends on that frozen weight.
  • The frame-level speaker posteriors SD-CTC produces could serve as a cheap source of pseudo diarization labels for unannotated conversation corpora, although the paper does not claim speaker tracking across segments.
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

3 major / 3 minor

Summary. The paper proposes Speaker-Distinguishable CTC (SD-CTC), an extension of CTC that jointly predicts a token and a speaker label for each frame, and integrates it with Serialized Output Training (SOT) via a multi-task loss. The SD-CTC loss is computed separately for each speaker using a speaker-specific blank token, and the method is trained with frozen single-speaker token posteriors and then fine-tuned on multi-talker mixtures. Experiments on LibriSpeechMix report that SOT + SD-CTC achieves 3.5% cpWER with CTC rescoring versus 4.7% for the SOT baseline, a 26% relative reduction, and matches methods that use auxiliary information. The paper also presents encoder visualizations suggesting improved speaker separation.

Significance. If the empirical results are robust, the paper makes a useful contribution: it shows that a lightweight, auxiliary-information-free multi-task loss can improve speaker attribution in SOT-based multi-talker ASR, with no increase in model parameters relative to the baseline and a clear ablation isolating the effect of SD-CTC. The algebraic derivation in Section 3 is clear and internally consistent, and the controlled comparison against SOT, SACTC, and GEncSep is a strength. However, the central empirical claim rests on a single evaluation run without variance estimates, and the method's core frame-factorization assumption is only tested on synthetic mixtures with substantial non-overlapped speech, leaving its applicability to real overlapping conversation unquantified.

major comments (3)
  1. [Section 3.1-3.3, Eq. (3)] The derivation of the speaker-specific token probability assumes that each frame belongs to exactly one speaker, but this is false for genuinely overlapping speech where both speakers are acoustically present. Equations (1) and (2) cannot assign probability mass to tokens from both speakers in the same frame; the speaker-specific blank term in Eq. (3) instead treats the non-target speaker's speech as if no token were emitted. The paper acknowledges this assumption in Section 3.1 but never quantifies its effect. Since the evaluation is on LibriSpeechMix, where mixtures are synthetic and often include long non-overlapped segments, the 26% cpWER improvement may reflect robustness to partial overlap rather than frame-level speaker distinction in fully overlapping speech. The authors should report performance stratified by overlap ratio, evaluate on a corpus with real overlapping speech, or otherwise provide evidence about how the assumption affects overlapping frames.
  2. [Section 4.2, Table 1] The central empirical claim of a 26% relative cpWER reduction is based on a single evaluation run with no error bars, confidence intervals, or significance tests. Given that the difference between SOT + SD-CTC (AED-only, 4.1%) and the SOT baseline (4.7%) is only 0.6 points absolute, it is important to show that the improvement is stable across multiple training seeds or random initializations. The authors should provide repeated runs or at least report the variance across evaluation subsets.
  3. [Section 3.4] The token posterior P_v is frozen from single-speaker pretraining and is used as a weight for speaker prediction on mixed audio. On overlapping frames, P_v was never trained to handle two simultaneous speakers, so it may systematically favor the louder or more prominent speaker, making the product P_s(σ|x_t)P_v(ρ|x_t) unreliable for the quieter speaker's correct token. The paper does not analyze this effect; an experiment that varies the signal-to-noise ratio per speaker or compares against a version where P_v is not frozen would help substantiate the claim that SD-CTC learns true speaker distinction rather than exploiting the dominant speaker's tokens.
minor comments (3)
  1. [Section 4.3] The sentence "same sample used in Figure 2.2" should refer to Figure 2(a), not "Figure 2.2."
  2. [Abstract and Section 4.2] The phrase "multi-task learning with SD-CTC and SOT reduces the error rate of the SOT model by 26%" should be clarified: the 26% figure is obtained with CTC rescoring in addition to multi-task training; the ablation with AED-only inference shows a smaller reduction from 4.7% to 4.1% (about 13% relative).
  3. [Section 3.2] The notation for the speaker-specific blank token is inconsistent: the text defines "<¬s>" but the formula in Eq. (2) and surrounding text sometimes refers to it with different spacing; a consistent notation would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SD-CTC auxiliary loss is trained and evaluated on held-out data, and no fitted input is relabeled as a prediction.

full rationale

The paper's derivation chain is self-contained and empirically grounded. SD-CTC defines an auxiliary CTC loss (Eqs. 3-5) from the acoustic features and the per-speaker transcriptions y_sigma, which are the same external supervision used by standard SOT; the loss is optimized by gradient descent and then evaluated on the held-out LibriSpeechMix test set. The claimed 26% relative cpWER reduction is a comparison between two trained systems (SOT Baseline vs. SOT+SD-CTC) on unseen data, not an identity or a quantity forced by construction. The one-frame-one-speaker factorization in Eq. (3) is an explicitly stated modeling assumption (Section 3.1), not a circular definition: it restricts the model family but does not encode the evaluation target, and the same assumption underlies standard CTC. Freezing the pre-trained token posterior P_v during fine-tuning (Section 3.4) is a training-pipeline choice; it does not fit any parameter to the cpWER result. No load-bearing self-citation is used: citations to prior CTC/attention and SOT work supply standard components rather than the claimed improvement, and no uniqueness theorem or ansatz is imported from the authors' own prior work. The paper also states a limitation that SD-CTC cannot assign multiple speakers to one frame (Section 4.2), which is a correctness/generalization risk for real overlapping speech, but that limitation is not a circularity because it does not make the empirical result equivalent to its inputs.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No free parameters are fitted to data in the derivation; the only hand-set numbers are training hyperparameters. The central modeling axioms are the single-speaker-per-frame assumption, the conditional independence in CTC, the factorization of speaker and token probabilities, and per-speaker transcript supervision.

free parameters (2)
  • SD-CTC loss weight = 0.3
    Hand-set weight for the SD-CTC auxiliary loss in multi-task training; no sensitivity analysis is reported.
  • CTC rescoring weight = 0.3
    Hand-set weight for the SD-CTC log-likelihood when rescoring beam-search hypotheses during inference; no sensitivity analysis is reported.
assumptions (4)
  • domain assumption CTC factorizes the label sequence probability as a product over time steps, assuming conditional independence of labels given the acoustic sequence.
    SD-CTC is a CTC loss (Section 3.3), so it inherits the per-frame independence assumption of standard CTC.
  • domain assumption Each audio frame belongs to exactly one speaker.
    Section 3.1 explicitly assumes one speaker per frame; this is violated by genuine overlap and is the main limitation for real conversations.
  • ad hoc to paper Speaker posterior and token posterior factor independently for each frame.
    Eq. (1) defines P(sigma,rho|x_t) as the product Ps(sigma|x_t)Pv(rho|x_t), a factorization introduced for SD-CTC.
  • domain assumption Per-speaker reference transcripts are available for every training segment.
    The SD-CTC loss (Eq. 4) requires the correct token sequence y_sigma for each speaker; this is standard supervised MT-ASR data, not auxiliary timing information.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Speaker-Distinguishable CTC: Learning Speaker Distinction Using CTC for Multi-Talker Speech Recognition." pith.science (2026). https://pith.science/paper/4JEBIBFJ

@misc{pith2026250607515,
  author       = {Pith},
  title        = {Pith review of: Speaker-Distinguishable CTC: Learning Speaker Distinction Using CTC for Multi-Talker Speech Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4JEBIBFJ}},
  note         = {Machine review of arXiv:2506.07515}
}
read the original abstract

This paper presents a novel framework for multi-talker automatic speech recognition without the need for auxiliary information. Serialized Output Training (SOT), a widely used approach, suffers from recognition errors due to speaker assignment failures. Although incorporating auxiliary information, such as token-level timestamps, can improve recognition accuracy, extracting such information from natural conversational speech remains challenging. To address this limitation, we propose Speaker-Distinguishable CTC (SD-CTC), an extension of CTC that jointly assigns a token and its corresponding speaker label to each frame. We further integrate SD-CTC into the SOT framework, enabling the SOT model to learn speaker distinction using only overlapping speech and transcriptions. Experimental comparisons show that multi-task learning with SD-CTC and SOT reduces the error rate of the SOT model by 26% and achieves performance comparable to state-of-the-art methods relying on auxiliary information.

Figures

Figures reproduced from arXiv: 2506.07515 by the authors.

Figure 1
Figure 1. Overview of proposed method. Our contributions are twofold: i) we develop SD-CTC, which jointly assigns token and speaker labels, and ii) we demonstrate its competitive performance on the Libri￾SpeechMix dataset without auxiliary information. These achievements contribute to improving the robustness of conver￾sational speech recognition. 2. Serialized Output Training Revisited 2.1. Original serialized output trainin… view at source ↗
Figure 2
Figure 2. a presents an attention map of the final decoder layer, obtained via teacher-forcing on a sample where the SOT model omitted one speaker’s transcription. When provided with oracle tokens, the final layer correctly aligns acoustic frames with tokens, indicating that misalignment is not the source of the error. Figure 2b shows an LDA visualization of encoder out￾puts from 100 LibriSpeechMix samples, where the LDA clas… view at source ↗
Figure 3
Figure 3. LDA visualization of encoder outputs attended by each token where green represents first speaker, orange repre￾sents second speaker, and purple represents <sc> token. (a) w/o SD-CTC (b) w/ SD-CTC [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Attention map of second decoder layers. Target utter￾ance is test-clean-2mix-1541 in LibriSpeechMix. whereas the SOT model attends to one speaker’s tokens, the model trained with the proposed method attends separately to each speaker’s tokens, effectively assigning fra…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 28 canonical work pages

  1. [1]

    Speaker-Distinguishable CTC: Learning Speaker Distinction Using CTC for Multi-Talker Speech Recognition

    Introduction Multi-talker automatic speech recognition (MT-ASR) is a tech- nique of transcribing conversations involving multiple speakers, often with overlapping speech, into separate text streams for each speaker. This technology has been extensively studied for applications such as meeting transcription and automatic subti- tle generation for talk show...

    work page Pith review arXiv 2025

  2. [2]

    Original serialized output training SOT is a method for MT-ASR using an AED model, as illus- trated on the left side of Figure 1

    Serialized Output Training Revisited 2.1. Original serialized output training SOT is a method for MT-ASR using an AED model, as illus- trated on the left side of Figure 1. For a set of utterances spoken by multiple speakers, the SOT model is trained to predict a to- ken sequence that concatenates the transcriptions of all speakers in the order they begin ...

    work page

  3. [3]

    Proposed Method 3.1. Speaker-Distinguishable CTC Speaker-Distinguishable CTC (SD-CTC) is a framework that enables the encoder to learn speaker-distinguishable represen- tations using only overlapping speech and their correspond- ing transcriptions. In contrast to prior CTC-based auxiliary losses for SOT [18, 20], SD-CTC avoids speaker extraction or frame ...

    work page

  4. [4]

    Auxiliary info

    Experiments We evaluate our method using ESPnet [21], training on Libri- Speech [22] and testing on LibriSpeechMix [3]. 4.1. Experimental setups The training procedure followed the approach in [7]. Specifi- cally, for each LibriSpeech sample, a second sample was ran- domly selected and mixed with a randomly determined delay. The number of mixed speech sam...

    work page

  5. [5]

    Conclusions In this paper, we propose Speaker-Distinguishable CTC, which improves multi-talker ASR accuracy using only overlapping speech and each speaker’s transcript. Our evaluation experi- ments demonstrate that multi-task learning with SOT and SD- CTC significantly improves recognition accuracy, achieving performance equivalent to state-of-the-art met...

    work page

  6. [6]

    Recognizing multi-talker speech with permutation invariant training,

    D. Yu, X. Chang, and Y . Qian, “Recognizing multi-talker speech with permutation invariant training,” inInterspeech 2017, 2017, pp. 2456–2460

    work page 2017

  7. [7]

    A purely end-to-end system for multi-speaker speech recognition,

    H. Seki, T. Hori, S. Watanabe, J. Le Roux, and J. R. Hershey, “A purely end-to-end system for multi-speaker speech recognition,” inProceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), I. Gurevych and Y . Miyao, Eds. Melbourne, Australia: Association for Computational Linguistics, Jul. 2018, pp. 2...

    work page 2018

  8. [8]

    Seri- alized output training for end-to-end overlapped speech recogni- tion,

    N. Kanda, Y . Gaur, X. Wang, Z. Meng, and T. Yoshioka, “Seri- alized output training for end-to-end overlapped speech recogni- tion,” inInterspeech 2020, 2020, pp. 2797–2801

    work page 2020

Show all 34 references
  1. [9]

    Alignment-free train- ing for transducer-based multi-talker asr,

    T. Moriya, S. Horiguchi, M. Delcroix, R. Masumura, T. Ashihara, H. Sato, K. Matsuura, and M. Mimura, “Alignment-free train- ing for transducer-based multi-talker asr,” inICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025, pp. 1–5

    2025

  2. [10]

    Empowering whisper as a joint multi-talker and target-talker speech recognition system,

    L. Meng, J. Kang, Y . Wang, Z. Jin, X. Wu, X. Liu, and H. Meng, “Empowering whisper as a joint multi-talker and target-talker speech recognition system,” inInterspeech 2024, 2024, pp. 4653– 4657

    2024

  3. [11]

    Listen, attend and spell: A neural network for large vocabulary conversational speech recognition,

    W. Chan, N. Jaitly, Q. Le, and O. Vinyals, “Listen, attend and spell: A neural network for large vocabulary conversational speech recognition,” in2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2016, pp. 4960–4964

    2016

  4. [12]

    Streaming multi-talker asr with token-level serialized output training,

    N. Kanda, J. Wu, Y . Wu, X. Xiao, Z. Meng, X. Wang, Y . Gaur, Z. Chen, J. Li, and T. Yoshioka, “Streaming multi-talker asr with token-level serialized output training,” inInterspeech 2022, 2022, pp. 3774–3778

    2022

  5. [13]

    Sa-sot: Speaker- aware serialized output training for multi-talker asr,

    Z. Fan, L. Dong, J. Zhang, L. Lu, and Z. Ma, “Sa-sot: Speaker- aware serialized output training for multi-talker asr,” inICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 9986–9990

    2024

  6. [14]

    Ba- sot: Boundary-aware serialized output training for multi-talker asr,

    Y . Liang, F. Yu, Y . Li, P. Guo, S. Zhang, Q. Chen, and L. Xie, “Ba- sot: Boundary-aware serialized output training for multi-talker asr,” inInterspeech 2023, 2023, pp. 3487–3491

    2023

  7. [15]

    Joint autoregressive modeling of end-to-end multi-talker over- lapped speech recognition and utterance-level timestamp predic- tion,

    N. Makishima, K. Suzuki, S. Suzuki, A. Ando, and R. Masumura, “Joint autoregressive modeling of end-to-end multi-talker over- lapped speech recognition and utterance-level timestamp predic- tion,” inInterspeech 2023, 2023, pp. 2913–2917

    2023

  8. [16]

    Improv- ing multi-speaker asr with overlap-aware encoding and monotonic attention,

    T. Li, F. Wang, W. Guan, L. Huang, Q. Hong, and L. Li, “Improv- ing multi-speaker asr with overlap-aware encoding and monotonic attention,” inICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 12 416–12 420

    2024

  9. [17]

    Somsred: Sequential output mod- eling for joint multi-talker overlapped speech recognition and speaker diarization,

    N. Makishima, N. Kawata, M. Ihori, T. Tanaka, S. Orihashi, A. Ando, and R. Masumura, “Somsred: Sequential output mod- eling for joint multi-talker overlapped speech recognition and speaker diarization,” inInterspeech 2024, 2024, pp. 1660–1664

    2024

  10. [18]

    End-to-end speaker-attributed asr with trans- former,

    N. Kanda, G. Ye, Y . Gaur, X. Wang, Z. Meng, Z. Chen, and T. Yoshioka, “End-to-end speaker-attributed asr with trans- former,” inInterspeech 2021, 2021, pp. 4413–4417

    2021

  11. [19]

    Speaker mask transformer for multi-talker overlapped speech recognition,

    P. Shen, X. Lu, and H. Kawai, “Speaker mask transformer for multi-talker overlapped speech recognition,” 2023. [Online]. Available: https://arxiv.org/abs/2312.10959

    2023 arXiv

  12. [20]

    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,” inInterspeech 2022, 2022, pp. 560–564

    2022

  13. [21]

    Hy- brid ctc/attention architecture for end-to-end speech recognition,

    S. Watanabe, T. Hori, S. Kim, J. R. Hershey, and T. Hayashi, “Hy- brid ctc/attention architecture for end-to-end speech recognition,” IEEE Journal of Selected Topics in Signal Processing, vol. 11, no. 8, pp. 1240–1253, 2017

    2017

  14. [22]

    Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks,

    A. Graves, S. Fern ´andez, F. Gomez, and J. Schmidhuber, “Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks,” inProceedings of the 23rd International Conference on Machine Learning, ser. ICML ’06. New York, NY , USA: Asso...

    2006

  15. [23]

    Serialized speech infor- mation guidance with overlapped encoding separation for multi- speaker automatic speech recognition,

    H. Shi, Y . Gao, Z. Ni, and T. Kawahara, “Serialized speech infor- mation guidance with overlapped encoding separation for multi- speaker automatic speech recognition,” in2024 IEEE Spoken Lan- guage Technology Workshop (SLT), 2024, pp. 193–199

    2024

  16. [24]

    Cross-speaker encoding network for multi-talker speech recog- nition,

    J. Kang, L. Meng, M. Cui, H. Guo, X. Wu, X. Liu, and H. Meng, “Cross-speaker encoding network for multi-talker speech recog- nition,” inICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 11 986–11 990

    2024

  17. [25]

    Disentangling speakers in multi-talker speech recogni- tion with speaker-aware ctc,

    J. Kang, L. Meng, M. Cui, Y . Wang, X. Wu, X. Liu, and H. Meng, “Disentangling speakers in multi-talker speech recogni- tion with speaker-aware ctc,” inICASSP 2025 - 2025 IEEE Inter- national Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025, pp. 1–5

    2025

  18. [26]

    Espnet: End-to-end speech pro- cessing toolkit,

    S. Watanabe, T. Hori, S. Karita, T. Hayashi, J. Nishitoba, Y . Unno, N. Enrique Yalta Soplin, J. Heymann, M. Wiesner, N. Chen, A. Renduchintala, and T. Ochiai, “Espnet: End-to-end speech pro- cessing toolkit,” inInterspeech 2018, 2018, pp. 2207–2211

    2018

  19. [27]

    Lib- rispeech: An asr corpus based on public domain audio books,

    V . Panayotov, G. Chen, D. Povey, and S. Khudanpur, “Lib- rispeech: An asr corpus based on public domain audio books,” in2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2015, pp. 5206–5210

    2015

  20. [28]

    Audio aug- mentation for speech recognition,

    T. Ko, V . Peddinti, D. Povey, and S. Khudanpur, “Audio aug- mentation for speech recognition,” inInterspeech 2015, 2015, pp. 3586–3589

    2015

  21. [29]

    Data augmentation for children’s speech recognition – the

    G. Chen, X. Na, Y . Wang, Z. Yan, J. Zhang, S. Ma, and Y . Wang, “Data augmentation for children’s speech recognition – the ”ethiopian” system for the slt 2021 children speech recognition challenge,” 2020. [Online]. Available: https: //arxiv.org/abs/2011.04547

    2021 arXiv

  22. [30]

    Specaugment: A simple data augmentation method for automatic speech recognition,

    D. S. Park, W. Chan, Y . Zhang, C.-C. Chiu, B. Zoph, E. D. Cubuk, and Q. V . Le, “Specaugment: A simple data augmentation method for automatic speech recognition,” inInterspeech 2019, 2019, pp. 2613–2617

    2019

  23. [31]

    Specaugment on large scale datasets,

    D. S. Park, Y . Zhang, C.-C. Chiu, Y . Chen, B. Li, W. Chan, Q. V . Le, and Y . Wu, “Specaugment on large scale datasets,” inICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 6879–6883

    2020

  24. [32]

    Conformer: Convolution-augmented transformer for speech recognition,

    A. Gulati, J. Qin, C.-C. Chiu, N. Parmar, Y . Zhang, J. Yu, W. Han, S. Wang, Z. Zhang, Y . Wu, and R. Pang, “Conformer: Convolution-augmented transformer for speech recognition,” in Interspeech 2020, 2020, pp. 5036–5040

    2020

  25. [33]

    Acoustic- to-word attention-based model complemented with character- level ctc-based model,

    S. Ueno, H. Inaguma, M. Mimura, and T. Kawahara, “Acoustic- to-word attention-based model complemented with character- level ctc-based model,” in2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018, pp. 5804–5808

    2018

  26. [34]

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

    S. Watanabe, M. Mandel, J. Barker, E. Vincent, A. Arora, X. Chang, S. Khudanpur, V . Manohar, D. Povey, D. Raj, D. Sny- der, A. S. Subramanian, J. Trmal, B. B. Yair, C. Boeddeker, Z. Ni, Y . Fujita, S. Horiguchi, N. Kanda, T. Yoshioka, and N. Ryant, “Chime-6 challenge: Tacklin...

    2020

Pith tools

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