Pith. sign in

REVIEW 4 major objections 5 minor 35 references

A two-layer module using only beam-search scores and ranks cuts ASR token-level worst-case calibration error from 20% to 4.5%.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 09:42 UTC pith:ANZCACR3

load-bearing objection Solid empirical case that beam-search score/rank features alone yield strong ASR confidence calibration; exact MCE numbers need a fixed-binning check but the core finding looks real. the 4 major comments →

arxiv 2607.29299 v1 pith:ANZCACR3 submitted 2026-07-31 eess.AS

Leveraging Beam Search Information for Confidence Estimation in E2E ASR

Yichen Jia , Hugo Van hamme This is my paper
classification eess.AS
keywords confidence estimationautomatic speech recognitionbeam searchcalibrationmaximum calibration errorend-to-end ASRRNN-Tarchitecture-agnostic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper tries to establish that the information already produced by beam-search decoding—token scores, ranks, and top-K alternatives—is enough to build well-calibrated confidence estimates for end-to-end speech recognition, without needing any internal model features. The proposed Score-Rank Confidence Estimation Module (SR-CEM) is a two-layer network of about 0.5k parameters that learns to map these decoding traces to confidence. On LibriSpeech test-clean it lowers token-level Maximum Calibration Error from 20.04% (softmax confidence) to 4.50% and Expected Calibration Error from 1.75% to 0.30%, with word-level MCE down from 17.91% to 8.17%. The authors show the same module remains effective when the ASR architecture changes (hybrid, attention-only, CTC-only, RNN-T), when the language switches to Dutch, and under noisy and conversational conditions, which matters because reliable confidence is what lets downstream systems decide whether to trust, flag, or correct a transcription.

Core claim

The paper's central claim is that the confidence signal an ASR system needs is already present in its own beam-search output. For each token in the winning hypothesis, SR-CEM reads the token's score, its rank in the per-step vocabulary score vector, the cumulative scores of the context before and after it, and the top-4 competing scores; for words it uses the summed token score, the maximum token rank, context sums, and the number of tokens. A single ReLU hidden layer and a sigmoid output are trained with binary cross-entropy against token/word correctness labels obtained by Levenshtein alignment. The authors explain the method's effectiveness through a local-global mismatch: softmax confide

What carries the argument

Score-Rank Confidence Estimation Module (SR-CEM): a two-layer MLP (one 64-unit ReLU hidden layer, sigmoid output) that maps beam-search-derived features to confidence. Token-level features: the selected token's score, its rank in the score vector, cumulative scores of preceding and succeeding tokens, and the top-4 scores; word-level features: summed word score, maximum token rank, cumulative context scores, and token count. The mechanism is that rank captures local competition among alternatives, context sums capture global hypothesis quality, and top-K scores capture uncertainty among alternatives—together they compensate for the mismatch between the local probability softmax reports and th

Load-bearing premise

The headline improvements over softmax assume the re-implemented baseline configurations (Xformer, E2EXformer) are faithful to the originals and that the adaptive-binning calibration metric does not systematically favor SR-CEM's output distribution; the method itself also assumes beam search explores enough alternatives for ranks and top-K scores to be informative.

What would settle it

Re-run the LibriSpeech test-clean evaluation with greedy decoding instead of beam search: if token-level MCE remains close to 4.50%, the claim that ranking information carries the signal would be refuted. Alternatively, recompute ECE and MCE with a fixed 15-bin equal-width binning; if the relative improvement over softmax largely disappears, the result is an artifact of the adaptive-binning metric rather than genuine calibration.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The same trained SR-CEM transfers across ASR architectures—hybrid CTC/attention, attention-only, CTC-only, and RNN-T—without re-engineering, because it never touches encoder, decoder, or attention states.
  • Worst-case calibration (MCE) improves by roughly 50–70% relative to softmax confidence across clean, noisy, conversational, and out-of-domain conditions, which is the property most relevant for deciding when to trust a transcription in production.
  • Deployment cost is near zero: the module has 0.4–0.6k parameters, stores about 12KB of features per 150-token utterance, and adds less than 0.1 ms inference latency since scores and ranks are already computed during decoding.
  • The method inherits the limitations of beam search: greedy decoding or aggressive pruning removes the ranking signal and degrades calibration, and severe train/test mismatch (clean-to-noisy, or new-domain vocabulary) still raises MCE to 19.59% out-of-domain.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The local-global mismatch mechanism is probably not confined to ASR: any autoregressive decoder that ranks hypotheses by cumulative score while emitting per-step probabilities—including large language models at inference—could show the same underconfidence for correct non-top-1 tokens, suggesting SR-CEM-style rank features may transfer to other sequence tasks.
  • Because the paper attributes the word-level performance gap (8.17% vs 4.50% MCE) to the absence of top-K features, a direct extension would be to include N-best list statistics at word level; such features might close the gap at modest computational cost.
  • SR-CEM's near-total reliance on beam-search traces implies that ASR toolkits could emit calibrated confidence as a standard decoding byproduct, turning confidence estimation from a post-hoc model into a free output—provided the decoding configuration keeps beam search informative.
  • A testable domain-adaptation probe: since the learned calibration curve is domain-specific (the paper states cross-language deployment requires retraining), one could evaluate how few target-domain utterances are needed to re-fit the 0.6k-parameter module and restore low MCE.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes SR-CEM, a confidence estimation module for end-to-end ASR that uses only beam-search score and rank information, rather than architecture-specific internal representations. The module is trained with binary cross-entropy on token/word correctness labels derived from Levenshtein alignment. The authors report extensive experiments across hybrid CTC/Attention, attention-only, CTC-only, and RNN-T models, on English (LibriSpeech, Common Voice, Libri-Adapt, CHiME-6) and Dutch (CGN), claiming consistent improvements in calibration, especially in Maximum Calibration Error, over softmax confidence and several architecture-dependent baselines. The headlining result is a token-level MCE of 4.50% and ECE of 0.30% on LibriSpeech test-clean, versus 20.04% and 1.75% for softmax confidence. The paper also includes an ablation study and a discussion of limitations.

Significance. If the results hold, the paper makes a useful contribution: it shows that simple, publicly available beam-search quantities can yield competitive confidence estimates without touching the internals of the ASR model, which is attractive for deployment across heterogeneous systems. The experimental scope is unusually broad, covering four decoding paradigms, two languages, and noisy/conversational conditions, and the authors are appropriately candid about limitations. The proposed module is also extremely light (sub-1k parameters), and the authors state that code will be released. The central claim is plausible and the paper presents a clear motivating analysis (Eq. 5) of the local-global mismatch in beam search. However, the reliability of the quantitative conclusions is currently weakened by (i) an underspecified adaptive-binning calibration metric, (ii) the absence of any variance or significance analysis for any reported number, and (iii) uncertainty about the fairness of the re-implemented baselines. These issues are addressable and do not appear to invalidate the core idea.

major comments (4)
  1. [Sec. II-B, Eqs. (3)-(4), Sec. V-K] The central MCE/ECE claims depend entirely on the 'adaptive binning' procedure, but the manuscript does not specify the algorithm: no bin count, no boundary criterion, no handling of ties or empty bins, and no indication of how K is selected. Because adaptive binning partitions the data differently for each score distribution, comparing MCE across methods is not a comparison on a fixed scale; a model can achieve low MCE simply because its scores induce few bins in poorly calibrated regions. The claimed 50-70% MCE reductions (Sec. V-K) are therefore not established unless the authors (1) fully specify the adaptive-binning algorithm, and (2) show the main comparisons also hold under a fixed binning scheme (e.g., equal-width or equal-frequency bins with a pre-specified K) for both ECE and MCE. This is load-bearing because the abstract's headline numbers are MCE values.
  2. [Tables 2-12] All reported metric values are single-run point estimates. No error bars, multiple seeds, confidence intervals, paired bootstrap intervals, or significance tests are provided anywhere in the results section. Consequently, the reader cannot distinguish real improvements from run-to-run variation, especially where differences are small, e.g., word-level ROC on LS (0.899 vs 0.931) or token-level CGN NCE (0.196 vs 0.259). This also affects the broad cross-architecture conclusion, since each cell is a single observation. The authors should provide variability estimates for at least the main comparisons (e.g., 3-5 training seeds or bootstrap over utterances) and state whether reported differences are stable.
  3. [Sec. IV-B, Tab. 1] The Xformer and E2EXformer baselines are compact variants selected by the authors because the original configurations 'were not released' (Sec. IV-B). The fairness of the comparison is therefore not established. A method that is under-tuned or under-capacity can make SR-CEM look better than it is relative to the original architecture-dependent approaches. The authors should report the hyperparameter search space and selection procedure for Xformer/E2EXformer, provide the resulting configurations, and ideally verify that the chosen sizes and training setup are competitive with the original papers. This is particularly important because the abstract claims 'significantly outperforming' softmax confidence, and the paper's stated advantage over baselines is part of that claim.
  4. [Sec. IV-C and Sec. V-D/V-E] The paper emphasizes architecture independence as the core contribution, yet the evaluations for attention-only and CTC-only decoding use the same backbone model with different decoding modes, and the RNN-T section uses a separately trained model. This is a reasonable design, but the paper does not explicitly discuss whether the relative gains of SR-CEM could be partly attributable to the fact that SR-CEM is the only method whose features are always extracted from the exact scoring procedure that determines correctness. For example, in the hybrid setting, correctness is determined by the final hypothesis combining CTC and attention scores, while baselines use only attention-based scoring; the paper acknowledges this in Sec. V-D but does not quantify how much of the advantage is explained by feature-source alignment. A short analysis of this confound, or at least a statement of its expect
minor comments (5)
  1. [Sec. V-D] The heading 'Experiemtn with Attention-Only Decoding' contains a typo ('Experiemtn' should be 'Experiment').
  2. [Sec. V-J] The phrase 'real-word setting' should be 'real-world setting'.
  3. [Table 11] In Table 11, the value '-0,416' uses a comma as a decimal separator, inconsistent with the rest of the tables that use periods. Please standardize.
  4. [References] Reference [15] is cited in the text as 'TruCLeS', and the authors mention 'TruCLeS' in the baseline description, but the reference title reads 'Teles: Temporal lexeme similarity score...'. Please verify the correct name and use it consistently in text and bibliography.
  5. [Sec. II-C, Eq. (5)] The bound c_softmax(y_t) <= 1/r(y_t) is a nice observation, but for clarity the authors should state explicitly that scores are assumed sorted in descending order before ranking and that ties are broken in a deterministic way, since the bound relies on the ordering of the top r scores.

Circularity Check

0 steps flagged

No significant circularity: SR-CEM is trained on external correctness labels from beam-search features, and its reported gains are empirical, not definitional.

full rationale

The derivation chain is self-contained. SR-CEM predicts confidence from beam-search features (Eqs. 9 and 14) using a small MLP trained with binary cross-entropy (Eq. 15) against correctness labels obtained from Levenshtein alignment between ground truth and hypothesis. These labels are external to the feature construction, so the output is not a restatement of the input. Eq. (5) is an explanatory upper bound on softmax confidence given rank; it is not used as a training signal or as a fitting constraint. The paper does not invoke a uniqueness theorem or rely on load-bearing self-citations; the cited prior work (e.g., [11], [5], [27]) is external or methodological. The main concerns—unspecified adaptive binning for ECE/MCE, single-run results, and re-implemented baselines for unreleased configurations—are threats to measurement validity and reproducibility, not circularity: they do not make any reported prediction equivalent to the model's inputs by construction. The central claim that beam-search scores/ranks alone can support competitive confidence estimation is an empirical result with independent test-set evaluation, so no circular step is established.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical or conceptual entities; the 'local-global mismatch' is an explanatory observation, not a postulated entity. The loaded assumptions are about validity of the experimental protocol: label generation via Levenshtein alignment, the adaptive-binning calibration metrics, the representative nature of the tested toolkits/architectures, and the faithfulness of baseline reimplementations.

free parameters (4)
  • top_k K = 4
    Number of alternative scores used in the token feature vector; chosen by the authors (Sec. III-A), not derived.
  • hidden dimension h = 64
    Hidden layer size of SR-CEM; recommended as 'balances expressiveness and simplicity' (Sec. IV-C).
  • word-rank aggregation = max
    Maximum token rank selected because it 'performed best in preliminary experiments' (Sec. III-B).
  • word-level score aggregation = mean
    Mean aggregation adopted from [11] (Sec. II-A); affects all word-level comparisons.
axioms (5)
  • standard math Softmax probability of a token is at most 1/rank (Eq. 5).
    Used to motivate rank as a confidence cue; follows directly from ranked score sums.
  • domain assumption Modern ASR toolkits perform beam search without prefix search and expose per-step score vectors in the winning hypothesis.
    Needed to make the RNN-T and CTC-only experiments meaningful (Sec. II-C, Sec. IV-D).
  • domain assumption Levenshtein alignment of recognized transcription to ground truth gives valid token/word correctness labels, with deletions excluded.
    All training targets and evaluation labels rest on this alignment (Sec. III-C).
  • domain assumption ECE/MCE computed via adaptive binning from [17] are valid calibration metrics as implemented by the authors.
    These are the core evaluation metrics; bin count and boundaries are data-dependent and not fully specified (Sec. II-B).
  • domain assumption The re-implemented baselines faithfully represent the respective original methods.
    Xformer/E2EXformer configurations were not released and were replaced by compact variants; TruCLeS settings were adopted with some dimension changes (Sec. IV-B).

pith-pipeline@v1.3.0-daily-deepseek · 14412 in / 13009 out tokens · 146817 ms · 2026-08-03T09:42:10.320016+00:00 · methodology

0 comments
read the original abstract

To estimate confidence for end-to-end Automatic Speech Recognition (ASR) systems, recent research has proposed Confidence Estimation Modules that incorporate features from the backbone ASR model. Most existing approaches, however, are architecture-dependent. In this paper, we propose the Score-Rank Confidence Estimation Module (SR-CEM), a lightweight module that leverages beam search information to generate token- and word-level confidence scores. Specifically, SR-CEM constructs features by combining the scores and ranks of tokens within a hypothesis. Experiments show that SR-CEM achieves effective calibration on both in-domain and out-of-domain English data. On the in-domain testset, it attains a Maximum Calibration Error of 4.50% and an Expected Calibration Error of 0.30% at the token level, significantly outperforming softmax confidence (20.04% and 1.75%, respectively). At the word level, SR-CEM achieves 8.17% and 0.35%, compared to 17.91% and 1.67% from softmax confidence. Furthermore, we demonstrate its robustness across hybrid and transducer ASR architectures with different decoding strategies, as well as on Dutch, noisy and conversational speech conditions. Our main finding is that SR-CEM is particularly effective in reducing Maximum Calibration Error, which is critical for reliable downstream use of ASR outputs, while maintaining architecture independence and generality across diverse evaluation conditions.

Figures

Figures reproduced from arXiv: 2607.29299 by Hugo Van hamme, Yichen Jia.

Figure 1
Figure 1. Figure 1: FIGURE 1: Reliability diagrams for softmax confidence [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIGURE 2: Reliability diagrams for SR-CEM confidence [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

35 extracted references · 8 linked inside Pith

  1. [1]

    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). IEEE, 2016, pp. 4960–4964

    2016

  2. [2]

    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 . Wuet al., “Conformer: Convolution-augmented transformer for speech recognition,”arXiv preprint arXiv:2005.08100, 2020

    Pith/arXiv arXiv 2005

  3. [3]

    A baseline for detecting misclassified and out-of-distribution examples in neural networks,

    D. Hendrycks and K. Gimpel, “A baseline for detecting misclassified and out-of-distribution examples in neural networks,” inProc. Inter- national Conference on Learning Representations (ICLR), 2017

    2017

  4. [4]

    On calibration of modern neural networks,

    C. Guo, G. Pleiss, Y . Sun, and K. Q. Weinberger, “On calibration of modern neural networks,” inInternational conference on machine learning. PMLR, 2017, pp. 1321–1330. 8 VOLUME , <Society logo(s) and publication title will appear here.>

    2017

  5. [5]

    Confidence estimation for attention-based sequence-to- sequence models for speech recognition,

    Q. Li, D. Qiu, Y . Zhang, B. Li, Y . He, P. C. Woodland, L. Cao, and T. Strohman, “Confidence estimation for attention-based sequence-to- sequence models for speech recognition,” inICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021, pp. 6388–6392

    2021

  6. [6]

    Active and semi-supervised learning in asr: Benefits on the acoustic and language models,

    T. Drugman, J. Pylkkonen, and R. Kneser, “Active and semi-supervised learning in asr: Benefits on the acoustic and language models,”arXiv preprint arXiv:1903.02852, 2019

    Pith/arXiv arXiv 1903

  7. [7]

    Active and semi-supervised learning for speech recogni- tion,

    F. Kreyssig, “Active and semi-supervised learning for speech recogni- tion,” Ph.D. dissertation, 2024

    2024

  8. [8]

    Posterior probability decoding, confidence estimation and system combination,

    G. Evermann and P. Woodland, “Posterior probability decoding, confidence estimation and system combination,” inProc. Speech Transcription Workshop, vol. 27. Baltimore, 2000, pp. 78–81

    2000

  9. [9]

    Towards interfacing large language models with ASR systems using confidence measures and prompting,

    M. Naderi, E. Hermann, A. Nanchen, S. Hovsepyan, and M. M. Doss, “Towards interfacing large language models with ASR systems using confidence measures and prompting,” inProc. Interspeech 2024, 2024, pp. 2980–2984

    2024

  10. [10]

    Beyond temperature scaling: Obtaining well-calibrated multi-class probabilities with dirichlet calibration,

    M. Kull, M. Perello Nieto, M. K ¨angsepp, T. Silva Filho, H. Song, and P. Flach, “Beyond temperature scaling: Obtaining well-calibrated multi-class probabilities with dirichlet calibration,”Advances in neural information processing systems, vol. 32, 2019

    2019

  11. [11]

    Learning word-level confidence for subword end-to-end asr,

    D. Qiu, Q. Li, Y . He, Y . Zhang, B. Li, L. Cao, R. Prabhavalkar, D. Bhatia, W. Li, K. Huet al., “Learning word-level confidence for subword end-to-end asr,” inICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021, pp. 6393–6397

    2021

  12. [12]

    Word-level confidence estimation for rnn transducers,

    M. Wang, H. Soltau, L. El Shafey, and I. Shafran, “Word-level confidence estimation for rnn transducers,” in2021 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). IEEE, 2021, pp. 1170–1177

    2021

  13. [13]

    Word-level confidence estimation for ctc models,

    B. Naowarat, T. Kongthaworn, and E. Chuangsuwanich, “Word-level confidence estimation for ctc models,” inProc. Interspeech 2023, 2023, pp. 3297–3301

    2023

  14. [14]

    Adopting whisper for confidence estimation,

    V . Aggarwal, S. S. Nair, Y . Verma, and Y . Jogi, “Adopting whisper for confidence estimation,” inICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    2025

  15. [15]

    Teles: Temporal lexeme similarity score to estimate confidence in end-to-end asr,

    N. Ravi, V . Aroraet al., “Teles: Temporal lexeme similarity score to estimate confidence in end-to-end asr,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2024

    2024

  16. [16]

    End-to-end speech recognition: A survey,

    R. Prabhavalkar, T. Hori, T. N. Sainath, R. Schl ¨uter, and S. Watanabe, “End-to-end speech recognition: A survey,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 325–351, 2023

    2023

  17. [17]

    Revisiting the evaluation of uncertainty estimation and its application to explore model complexity- uncertainty trade-off,

    Y . Ding, J. Liu, J. Xiong, and Y . Shi, “Revisiting the evaluation of uncertainty estimation and its application to explore model complexity- uncertainty trade-off,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020, pp. 4–5

    2020

  18. [18]

    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,” in2015 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2015, pp. 5206–5210

    2015

  19. [19]

    Hybrid ctc/attention architecture for end-to-end speech recognition,

    S. Watanabe, T. Hori, S. Kim, J. R. Hershey, and T. Hayashi, “Hybrid 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

  20. [20]

    ESPnet: End-to-end speech processing toolkit,

    S. Watanabe, T. Hori, S. Karita, T. Hayashi, J. Nishitoba, Y . Unno, N. E. Y . Soplin, J. Heymann, M. Wiesner, N. Chenet al., “ESPnet: End-to-end speech processing toolkit,” inProc. Interspeech 2018, 2018, pp. 2207–2211

    2018

  21. [21]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, “Pytorch: An imperative style, high-performance deep learning library,”arXiv preprint arXiv:1912.01703, 2019

    Pith/arXiv arXiv 1912

  22. [22]

    Com- mon voice: A massively-multilingual speech corpus,

    R. Ardila, M. Branson, K. Davis, M. Henretty, M. Kohler, J. Meyer, R. Morais, L. Saunders, F. M. Tyers, and G. Weber, “Com- mon voice: A massively-multilingual speech corpus,”arXiv preprint arXiv:1912.06670, 2019

    Pith/arXiv arXiv 1912

  23. [23]

    Robust speech recognition via large-scale weak super- vision,

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

    2023

  24. [24]

    Reproducing whisper-style training using an open-source toolkit and publicly available data,

    Y . Peng, J. Tian, B. Yan, D. Berrebbi, X. Chang, X. Li, J. Shi, S. Arora, W. Chen, R. Sharmaet al., “Reproducing whisper-style training using an open-source toolkit and publicly available data,” in2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). IEEE, 2023, pp. 1–8

    2023

  25. [25]

    Libri-adapt: a new speech dataset for unsupervised domain adaptation,

    A. Mathur, F. Kawsar, N. Berthouze, and N. D. Lane, “Libri-adapt: a new speech dataset for unsupervised domain adaptation,” inICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2020, pp. 7439–7443

    2020

  26. [26]

    The spoken dutch corpus. overview and first evaluation

    N. Oostdijket al., “The spoken dutch corpus. overview and first evaluation.” inLREC. Athens, Greece, 2000, pp. 887–894

    2000

  27. [27]

    Leveraging broadcast media subtitle transcripts for automatic speech recognition and subtitling,

    J. Ponceletet al., “Leveraging broadcast media subtitle transcripts for automatic speech recognition and subtitling,”arXiv preprint arXiv:2502.03212, 2025

    Pith/arXiv arXiv 2025

  28. [28]

    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. Rajet al., “Chime-6 challenge: Tackling multispeaker speech recognition for unsegmented recordings,” inCHiME 2020-6th International Workshop on Speech Processing in Everyday Environments, 2020

    2020

  29. [29]

    Adam: A method for stochastic optimiza- tion,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimiza- tion,” inProc. International Conference on Learning Representations (ICLR), 2015

    2015

  30. [30]

    Sequence transduction with recurrent neural networks,

    A. Graves, “Sequence transduction with recurrent neural networks,” in Proc. ICML Workshop on Representation Learning, 2012

    2012

  31. [31]

    Speechbrain: A general-purpose speech toolkit,

    M. Ravanelli, T. Parcollet, P. Plantinga, A. Rouhe, S. Cornell, L. Lu- gosch, C. Subakan, N. Dawalatabad, A. Heba, J. Zhonget al., “Speechbrain: A general-purpose speech toolkit,”arXiv preprint arXiv:2106.04624, 2021

    Pith/arXiv arXiv 2021

  32. [32]

    Nemo: a toolkit for building ai applications using neural modules,

    O. Kuchaiev, J. Li, H. Nguyen, O. Hrinchuk, R. Leary, B. Ginsburg, S. Kriman, S. Beliaev, V . Lavrukhin, J. Cooket al., “Nemo: a toolkit for building ai applications using neural modules,”arXiv preprint arXiv:1909.09577, 2019

    Pith/arXiv arXiv 1909

  33. [33]

    fairseq: A fast, extensible toolkit for sequence model- ing,

    M. Ott, S. Edunov, A. Baevski, A. Fan, S. Gross, N. Ng, D. Grangier, and M. Auli, “fairseq: A fast, extensible toolkit for sequence model- ing,”arXiv preprint arXiv:1904.01038, 2019

    Pith/arXiv arXiv 1904

  34. [34]

    Exploring architectures, data and units for streaming end-to-end speech recognition with rnn- transducer,

    K. Rao, H. Sak, and R. Prabhavalkar, “Exploring architectures, data and units for streaming end-to-end speech recognition with rnn- transducer,” in2017 IEEE automatic speech recognition and under- standing workshop (ASRU). IEEE, 2017, pp. 193–199

    2017

  35. [35]

    Improved estimation, evalua- tion and applications of confidence measures for speech recognition

    M.-h. Siu, H. Gish, and F. Richardson, “Improved estimation, evalua- tion and applications of confidence measures for speech recognition.” inEurospeech, 1997, pp. 831–834. Yichen Jia(Member, IEEE) received the B.S. degree in financial engineering from Sichuan Uni- versity, Chengdu, China, in 2021, and the M.S. degree in statistics and data science from KU...

    1997