Pith. sign in

REVIEW 3 major objections 7 minor 26 references

Related-Language Transfer Beats Baseline for Dhivehi ASR

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 →

Continual pre-training on Sinhala followed by Dhivehi fine-tuning achieves 12.89% WER, but KenLM decoding contributes 27+ absolute WER points, dwarfing the 0.61-point transfer gain.

T0 review reviewed 2026-07-08 challenge →

load-bearing objection Sinhala-to-Dhivehi transfer study: useful empirical work on an unstudied language pair, but the headline CPT result is architecturally confounded. the 3 major comments →

arxiv 2607.06289 v1 pith:M6YBEVZW submitted 2026-07-07 cs.CL

From Sinhala to Dhivehi: Cross-Lingual Transfer Learning for Low-Resource Speech Recognition

classification cs.CL
keywords automatic speech recognitioncross-lingual transfer learninglow-resource languagesDhivehiSinhalacontinual pre-trainingmultilingual fine-tuninglanguage model decoding
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.

The reading

This paper tests whether speech recognition for Dhivehi, a language spoken by fewer than half a million people with only 37 hours of validated training data, can be improved by leveraging Sinhala, a related Indo-Aryan language with roughly six times more available data. The authors run seventeen experiments across five transfer strategies—Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pre-training, and a Turkish control. Their best system, continual pre-training on Sinhala followed by Dhivehi fine-tuning with an external language model, achieves 12.89% word error rate, improving on the Dhivehi-only baseline of 13.50%. A Turkish control experiment, using an unrelated language matched in data volume, performs worse than the baseline, which the authors interpret as evidence that the gains stem from genuine linguistic relatedness rather than generic data augmentation. However, the paper also finds that the external n-gram language model contributes by far the largest improvement—roughly 28 absolute WER points across all systems—dwarfing every transfer strategy. Additionally, language identification tokens, often beneficial in large-scale multilingual ASR, hurt performance in this bilingual setting, and sequential fine-tuning failed to improve over the baseline at all.

Core claim

The central finding is that cross-lingual transfer from a related language to a low-resource one works, but only through specific training strategies—continual pre-training and multilingual fine-tuning without language ID tokens—while the choice of decoding configuration (an external n-gram language model) has a far larger effect on final accuracy than any transfer approach. The Turkish control experiment, which degrades performance below the baseline, is used to argue that the modest gains from Sinhala reflect genuine phonological and acoustic overlap between the two Insular Indo-Aryan languages rather than a generic benefit from adding more training data.

What carries the argument

The paper compares five transfer learning paradigms applied to self-supervised speech models (Wav2Vec2-BERT and XLS-R): (1) Dhivehi-only fine-tuning as baseline, (2) sequential fine-tuning from Sinhala to Dhivehi, (3) multilingual joint fine-tuning with and without language ID tokens, (4) continual pre-training on Sinhala audio followed by Dhivehi fine-tuning, and (5) a Turkish control replacing Sinhala with an unrelated language at matched data volume. An external 5-gram KenLM language model is integrated via shallow fusion with pyctcdecode beam search decoding (beam width 64, alpha 0.5, beta 1.0).

Load-bearing premise

The continual pre-training experiment uses a different base model (XLS-R) than all other experiments (Wav2Vec2-BERT), so the paper's claim that CPT is the best transfer strategy is confounded—XLS-R may simply be a stronger architecture independent of the transfer approach.

What would settle it

If replacing Sinhala with Turkish in the continual pre-training setup still produced WER improvements over the Dhivehi-only baseline comparable to the Sinhala CPT result, the claim that gains stem from linguistic relatedness would be undermined.

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

If this is right

  • For low-resource ASR, external language model decoding may matter more than the choice of acoustic model or transfer strategy, suggesting that n-gram LM integration should be a first-class experimental variable rather than a post-hoc addition.
  • Language ID tokens, standard in large-scale multilingual ASR, can be counterproductive in bilingual low-resource settings—implying that design choices from high-resource regimes require empirical validation before transfer to low-resource scenarios.
  • The saturation of multilingual benefit at modest source-language volumes (30 vs 60 hours of Sinhala yielding nearly identical Dhivehi WER) suggests that even small amounts of related-language data may suffice for cross-lingual transfer.
  • The Turkish control methodology—using an unrelated language at matched data volume—provides a template for distinguishing linguistic relatedness effects from generic data augmentation in other transfer learning studies.

Where Pith is reading between the lines

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

  • The architectural inconsistency between continual pre-training (XLS-R) and all other experiments (Wav2Vec2-BERT) means the best result cannot be cleanly attributed to the transfer strategy alone; a controlled comparison using the same base architecture for CPT vs. other strategies would be needed to isolate the effect.
  • The near-clustering of all systems at 40-44% WER without language model decoding suggests that the acoustic models alone may not be learning sufficiently discriminative representations in this data regime, and the KenLM is compensating for acoustic model weakness rather than complementing strong representations.
  • If the finding about language ID tokens generalizes, it could imply that explicit language conditioning is only beneficial when the model must distinguish among many languages; in a two-language setting, the added tokens may partition the decoder's capacity unnecessarily.
  • The 0.51 WER gap between the best transfer system (12.89%) and the Turkish control (13.77%) is small relative to the 28-point KenLM effect, raising the question of whether the relatedness signal would survive statistical significance testing across multiple random seeds or data splits.
Share X Bluesky LinkedIn Reddit HN

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

3 major / 7 minor

Summary. This paper investigates cross-lingual transfer learning from Sinhala to Dhivehi for low-resource ASR, conducting 17 experiments across five transfer paradigms (Dhivehi-only baseline, sequential fine-tuning, multilingual fine-tuning, continual pre-training, and a Turkish control). The strongest system—continual pre-training (CPT) on Sinhala followed by Dhivehi fine-tuning with KenLM decoding—achieves 12.89% WER and 2.70% CER, improving over the Dhivehi-only baseline (13.50% WER, 3.02% CER). The Turkish control experiment provides evidence that gains stem from linguistic relatedness rather than generic data augmentation. The study also finds that KenLM decoding is the single largest contributor to performance, that language ID tokens are harmful in this bilingual setting, and that multilingual fine-tuning without LID tokens is a competitive alternative to CPT.

Significance. The paper addresses a genuinely under-studied language pair (Sinhala–Dhivehi) and provides a controlled experimental comparison across multiple transfer paradigms, which is valuable for the low-resource ASR community. The inclusion of a Turkish unrelated-language control is a methodological strength, as is the finding that language ID tokens can be harmful in bilingual low-resource settings—a practically useful and falsifiable result. The reproducible code repository and the reporting of both WER and CER across all configurations are commendable. The KenLM effect analysis (Table IV) is a useful contribution showing that decoding configuration is a first-order effect. However, the central claim that CPT is the best transfer strategy is confounded by an architectural inconsistency (XLS-R for CPT vs. Wav2Vec2-BERT for all other experiments), and the transfer gains (0.24–0.61 absolute WER points) are reported without variance estimates, limiting the strength of the comparative conclusions.

major comments (3)
  1. §III.D.5 and §V.E: The CPT experiment uses XLS-R while all other experiments use Wav2Vec2-BERT. The paper acknowledges this as an 'architectural inconsistency' but does not address its implications for the central claim. The 0.61 WER improvement of CPT over the Dhivehi-only baseline cannot be cleanly attributed to the transfer strategy because XLS-R (pretrained on 128 languages, ~436K hours) may be a stronger base model than Wav2Vec2-BERT regardless of transfer. This is load-bearing for the claim that CPT is the best transfer strategy. At minimum, the paper should explicitly discuss this confound in the results interpretation (§V.A) rather than only in the limitations, and ideally should include a Dhivehi-only XLS-R baseline to isolate the architecture effect from the transfer effect.
  2. §III.D.6 and §IV.E: The Turkish control experiment is conducted only under the multilingual fine-tuning paradigm, not under CPT. The paper's claim that 'linguistic relatedness' drives the CPT improvement is therefore supported only indirectly, via a different transfer paradigm with a different architecture. The headline CPT result has neither an architectural control (same base model without Sinhala transfer) nor a relatedness control (CPT with Turkish). This should be explicitly acknowledged as a limitation of the relatedness claim, or a CPT Turkish control should be added.
  3. §V.E and Table III: All results are from single runs with no repeated runs, no cross-validation, and no multiple data splits. The transfer gains are small (0.24–0.61 absolute WER points). In low-resource ASR, single-run variance of this magnitude is plausible. Without any variance estimates or confidence intervals, it is unclear whether any transfer strategy meaningfully outperforms the baseline. The paper should either provide variance estimates from repeated runs or explicitly temper the comparative claims to reflect this uncertainty.
minor comments (7)
  1. Abstract: The second half is near-duplicated, repeating the list of five paradigms and the best-system description almost verbatim. This should be condensed.
  2. §III.B: The text says 'Wav2Vec was selected as the primary ASR model' but the experiments use Wav2Vec2-BERT (and XLS-R for CPT). The model naming should be stated consistently and precisely.
  3. §III.C: The KenLM hyperparameters (α=0.5, β=1.0, beam=64) are described as 'empirically tuned' but the search space and tuning procedure are not reported. A brief description of the tuning range and selection criterion would improve reproducibility.
  4. Table III: The 'Cont. pretrain Si→Dv' row should note the base model (XLS-R) explicitly in the table or its caption, so the architectural difference is visible to readers without consulting §V.E.
  5. §II.A: The text mentions 'XLS-R [8] extended this framework across 53 languages, while XLS-R [7] scaled to 128 languages.' The first appears to refer to XLSR-53 (Conneau et al., 2021) and the second to XLS-R (Babu et al., 2022). The naming should distinguish XLSR-53 from XLS-R to avoid confusion.
  6. §IV.G: The comparison with published benchmarks (Table VIII) notes that Ahmed [12] used Common Voice 13.0 while this work uses Common Voice 22.0. This dataset-version difference should be stated more prominently, as it affects the comparability of the 13.50% vs. 14.26% WER numbers.
  7. Figure 1 is referenced but not visible in the manuscript text provided. Ensure the figure clearly shows the genealogical relationship as described.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a careful and constructive review. The referee raises three substantive concerns, all of which are valid in whole or in part. We address each below and describe the revisions we will make.

read point-by-point responses
  1. Referee: CPT uses XLS-R while all other experiments use Wav2Vec2-BERT, confounding the claim that CPT is the best transfer strategy. The paper should discuss this in results interpretation, not just limitations, and ideally add a Dhivehi-only XLS-R baseline.

    Authors: The referee is correct that the architectural inconsistency between XLS-R (used for CPT) and Wav2Vec2-BERT (used for all other experiments) confounds the attribution of the CPT improvement to the transfer strategy rather than to the base model. We acknowledge this without reservation. In the current manuscript, this issue is mentioned only in §V.E (Limitations), which is insufficient given that the headline result depends on it. We will make two changes. First, we will add an explicit discussion of this confound in §V.A (Transfer Strategy Matters More Than Relatedness Alone), making clear that the 0.61 absolute WER improvement of CPT over the Dhivehi-only baseline cannot be cleanly separated from the potentially stronger base architecture. We will temper the claim from 'CPT is the best transfer strategy' to 'CPT yields the best observed performance in this study, though the architectural difference between XLS-R and Wav2Vec2-BERT prevents a clean isolation of the transfer effect from the architecture effect.' Second, regarding the Dhivehi-only XLS-R baseline: we agree this is the ideal control and will attempt to run it. However, we must be honest that computational constraints may prevent us from completing this additional experiment within the revision timeframe. If we cannot add it, we will state explicitly that this control is missing and that the CPT result should be interpreted with the architectural confound in mind. We will not claim a clean attribution that the data do not support. revision: partial

  2. Referee: The Turkish control is conducted only under multilingual fine-tuning, not CPT. The relatedness claim for CPT is therefore only indirectly supported. This should be acknowledged or a CPT Turkish control should be added.

    Authors: This is a fair and accurate observation. The Turkish control experiment (§III.D.6, §IV.E) is conducted under the multilingual fine-tuning paradigm using Wav2Vec2-BERT, not under CPT using XLS-R. Therefore, the claim that linguistic relatedness drives the CPT improvement is supported only indirectly—via a different transfer paradigm and a different architecture. The manuscript's conclusion states that 'the Turkish control experiment confirms that observed improvements stem from linguistic relatedness,' which overstates what the evidence supports specifically for the CPT result. We will revise this in two ways. First, we will narrow the relatedness claim in the Conclusion and in §V.D to apply specifically to the multilingual fine-tuning paradigm, where the Turkish control is directly applicable. Second, we will add an explicit statement that the CPT result lacks both an architectural control (Dhivehi-only XLS-R) and a relatedness control (CPT with Turkish), and that the extension of the relatedness finding to CPT is an inference, not a directly tested claim. As with the XLS-R baseline, we will attempt to run a CPT Turkish control if computational resources permit, but we commit to the textual correction regardless of whether the additional experiment is completed. revision: partial

  3. Referee: All results are from single runs with no variance estimates. Transfer gains are small (0.24–0.61 absolute WER points). Without variance estimates, it is unclear whether any transfer strategy meaningfully outperforms the baseline.

    Authors: The referee is correct. The transfer gains are small—0.24 absolute WER for multilingual fine-tuning without LID tokens and 0.61 absolute WER for CPT—and these are from single runs with no repeated runs, cross-validation, or multiple data splits. In low-resource ASR with ~37 hours of training data, run-to-run variance of this magnitude is entirely plausible, and we cannot rule out that some or all of the observed differences are within noise. We will address this in two ways. First, we will add explicit hedging language throughout the results and discussion sections, stating that the observed differences are small and that without variance estimates we cannot confirm statistical significance. We will reframe the comparative claims as observed trends rather than established differences. Second, we will attempt to run at least 3–5 repeated runs for the key configurations (Dhivehi-only baseline, multilingual no-LID, and CPT) to provide standard deviations or confidence intervals. If we are able to complete these, we will report them in a revised Table III. If computational constraints prevent sufficient repeated runs, we will state this limitation explicitly and further temper the claims. We agree with the referee that the current framing is too strong for single-run results with sub-1-point differences. revision: partial

Circularity Check

0 steps flagged

No circularity found; the paper's central claims are empirical results from controlled experiments, not definitions or fitted constants renamed as predictions.

full rationale

The paper reports empirical ASR results from seventeen controlled experiments across five transfer-learning paradigms. The KenLM decoding equation (Eq. 1) is a standard shallow-fusion formula with fixed hyperparameters (α=0.5, β=1.0), not a quantity fitted to the target data and then presented as a prediction. The WER/CER metrics (Eq. 2) are standard evaluation formulas. The headline result (12.89% WER for CPT) is an empirical measurement, not a derived quantity that reduces to its inputs by construction. The Turkish control experiment provides an independent check on the relatedness claim. Self-citation is minimal—references [13] and [14] are cited for motivation but are not load-bearing for any derivation. The architectural inconsistency (XLS-R for CPT vs. Wav2Vec2-BERT for other experiments) is a confounding concern for correctness and attribution, but it is not circularity: the CPT result is still an empirical measurement, not a quantity defined in terms of the claim it supports. No step in the paper's chain reduces to its own inputs by construction.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 0 invented entities

No new entities are invented. The free parameters are standard ASR hyperparameters but are empirically set without systematic search. The domain assumptions are linguistically motivated but the key acoustic-overlap assumption is unmeasured. The single-run axiom is the most fragile: with transfer gains of 0.24-0.61 WER, run-to-run variance could change conclusions.

free parameters (6)
  • KenLM α (LM weight) = 0.5
    Empirically tuned; no search space or selection criterion reported (§III.C)
  • KenLM β (word bonus) = 1.0
    Empirically tuned; no search space or selection criterion reported (§III.C)
  • Beam width = 64
    Selected after testing 128; no systematic search reported (§III.C)
  • Sinhala subsample sizes = 30h, 60h
    Chosen to reduce source dominance; no principled selection criterion (§III.A)
  • Learning rate = 5e-5
    Standard value; not tuned for this task (§III.E)
  • Number of epochs = 10
    Fixed across all experiments; no early stopping or validation-based selection (§III.E)
axioms (4)
  • domain assumption Dhivehi and Sinhala share phonological and acoustic features that enable cross-lingual transfer
    Invoked in §I and §V.D; supported by comparative linguistics (Gnanadesikan 2016) but the acoustic overlap is assumed, not measured
  • domain assumption Character-level tokenisation is appropriate for Sinhala and Dhivehi ASR
    §III.B; justified by limited data and script differences, but contradicts Ahmed [12] who found subword modelling outperforms character vocabulary for Dhivehi
  • domain assumption Turkish is a valid unrelated-language control for isolating linguistic relatedness effects
    §III.D; reasonable but Turkish differs from Sinhala in data quality, recording conditions, and phonetic inventory, which could confound the comparison
  • ad hoc to paper Single-run results without cross-validation are sufficient for comparing transfer strategies
    §V.E acknowledges no repeated runs or cross-validation; with 0.24 WER differences between conditions, single-run variance could reverse rankings

reviewed 2026-07-08 · how reviews work

0 comments
Cite this review

Pith. "Pith review of From Sinhala to Dhivehi: Cross-Lingual Transfer Learning for Low-Resource Speech Recognition." pith.science (2026). https://pith.science/paper/M6YBEVZW

@misc{pith2026260706289,
  author       = {Pith},
  title        = {Pith review of: From Sinhala to Dhivehi: Cross-Lingual Transfer Learning for Low-Resource Speech Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M6YBEVZW}},
  note         = {Machine review of arXiv:2607.06289}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Dhivehi, the national language of the Maldives, is currently under-resourced for automatic speech recognition (ASR) and other NLP tasks. This study investigates whether cross-lingual transfer learning from Sinhala, a linguistically related, relatively well-resourced Insular Indo-Aryan language, can improve Dhivehi ASR. We conduct seventeen experiments across five transfer learning paradigms: Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pre-training, and a control using Turkish as an unrelated language. The strongest system, continual pre-training on Sinhala followed by fine-tuning on Dhivehi with KenLM, achieves 12.89% WER and 2.70% CER, outperforming the Dhivehi-only baseline by 13.50% WER and 3.02% CER. However, the adaptation strategy and decoding configuration are equally critical for a successful transfer learning experiment. We conduct seventeen controlled experiments spanning five transfer learning paradigms: Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pre-training, and a control experiment using Turkish as an unrelated language. The strongest system, continual pre-training on Sinhala followed by fine-tuning on Dhivehi with KenLM, achieves 12.89% WER and 2.70% CER, outperforming the Dhivehi-only baseline by 13.50% WER and 3.02% CER. The Turkish control experiment confirms that observed improvements stem from linguistic relatedness; adaptation strategy and decoding configuration are also critical.

Figures

Figures reproduced from arXiv: 2607.06289 by Lukmal Ilyas, Nevidu Jayatilleke.

Figure 1
Figure 1. Figure 1: Genealogical relationship of Sinhala, Dhivehi, and Turkish. [PITH_FULL_IMAGE:figures/full_fig_p002_1.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

26 extracted references · 26 canonical work pages · 1 internal anchor

  1. [1]

    Libri-light: A benchmark for asr with limited or no supervision,

    J. Kahn et al., “Libri-light: A benchmark for asr with limited or no supervision,” inProc. IEEE ICASSP, 2020, pp. 7669–7673

  2. [2]

    Scaling speech technology to 1,000+ languages,

    V . Pratap et al., “Scaling speech technology to 1,000+ languages,”J. Mach. Learn. Res., vol. 25, no. 97, pp. 1–52, 2024. [Online]. Available: http://jmlr.org/papers/v25/23-1318.html

  3. [3]

    The constitution of the republic of maldives,

    Republic of Maldives, “The constitution of the republic of maldives,” 2008

  4. [4]

    A. E. Gnanadesikan,Dhivehi: The language of the Maldives. Walter de Gruyter GmbH & Co KG, 2016, vol. 3

  5. [5]

    Common voice: A massively-multilingual speech corpus,

    R. Ardila et al., “Common voice: A massively-multilingual speech corpus,” inProc. 12th Lang. Resources Evaluation Conf. (LREC), Marseille, France, 2020, pp. 4218–4222

  6. [6]

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

    A. Baevski, Y . Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A framework for self-supervised learning of speech representations,” inAdvances in Neural Information Processing Systems, H. Larochelle, M. Ranzato, R. Hadsell, M. Balcan, and H. Lin, Eds., vol. 33. Curran Associates, Inc., 2020, pp. 12 449– 12 460. [Online]. Available: https://proceedings.neuri...

  7. [7]

    Xls-r: Self-supervised cross-lingual speech representa- tion learning at scale,

    A. Babu et al., “Xls-r: Self-supervised cross-lingual speech representa- tion learning at scale,” inProc. Interspeech, 2022, pp. 2278–2282

  8. [8]

    Unsupervised cross-lingual representation learning for speech recognition,

    A. Conneau et al., “Unsupervised cross-lingual representation learning for speech recognition,” inProc. Interspeech, 2021, pp. 2426–2430

  9. [9]

    Adapting multilingual speech representation model for a new, underresourced language through multilingual fine-tuning and continued pretraining,

    K. Nowakowski, M. Ptaszynski, K. Murasaki, and J. Nieuwa ˙zny, “Adapting multilingual speech representation model for a new, underresourced language through multilingual fine-tuning and continued pretraining,”Information Processing & Management, vol. 60, no. 2, p. 103148, Mar. 2023. [Online]. Available: https://doi.org/10.1016/j.ipm. 2022.103148

  10. [10]

    Adaptation techniques for low-resource asr,

    Y . Getman et al., “Adaptation techniques for low-resource asr,” inProc. Interspeech, 2024, pp. 2539–2543

  11. [11]

    Crowd-sourced speech corpora for multiple languages,

    O. Kjartansson et al., “Crowd-sourced speech corpora for multiple languages,” inProc. SLTU, Gurugram, India, 2018, pp. 52–55

  12. [12]

    Improving dhivehi automatic speech recognition,

    A. Ahmed, “Improving dhivehi automatic speech recognition,” inProc. ICNLSP, 2023, pp. 256–265

  13. [13]

    Dialect matters: Cross-lingual asr transfer for low- resource indic languages,

    A. Dhasmana et al., “Dialect matters: Cross-lingual asr transfer for low- resource indic languages,”arXiv preprint arXiv:2601.04373, 2026

  14. [14]

    Role of language relatedness in multilingual fine-tuning of language models: A case study in indo-aryan languages,

    T. I. Dhamecha et al., “Role of language relatedness in multilingual fine-tuning of language models: A case study in indo-aryan languages,” inProc. EMNLP, 2021, pp. 8584–8595

  15. [15]

    Robust speech recognition via large-scale weak supervision,

    A. Radford et al., “Robust speech recognition via large-scale weak supervision,” inProc. ICML, 2023, pp. 28 492–28 518

  16. [16]

    Improving sinhala speech recognition through e2e lf-mmi model,

    B. Gamage, T. Nadungodage, and R. Weerasinghe, “Improving sinhala speech recognition through e2e lf-mmi model,” inProc. ICON, 2021, pp. 213–219

  17. [17]

    Applicability of end-to-end deep neural architecture to sinhala speech recognition,

    B. Gamage, R. Pushpananda, T. Nadungodage, and R. Weerasinghe, “Applicability of end-to-end deep neural architecture to sinhala speech recognition,”Int. J. Adv. ICT Emerg. Regions, vol. 17, no. 1, 2024

  18. [18]

    Transfer learning for sinhala speech recognition,

    L. Nanayakkara and R. Weerasinghe, “Transfer learning for sinhala speech recognition,” inMachine Learning, Deep Learning and Com- putational Intelligence for Wireless Communication, 2024, pp. 17–28

  19. [19]

    Cross-language end-to-end speech recognition for low- resource tujia language,

    C. Yu et al., “Cross-language end-to-end speech recognition for low- resource tujia language,”Symmetry, vol. 11, no. 2, p. 179, 2019

  20. [20]

    Cross-lingual transfer learning for low-resource cantonese speech recognition,

    X. Li, “Cross-lingual transfer learning for low-resource cantonese speech recognition,” 2024

  21. [21]

    Energy barriers for boundary nucleation in a two-well model without gauge invariance

    L. G. Pillai et al., “Multistage fine-tuning strategies for low-resource asr,”arXiv preprint arXiv:2403.04567, 2024

  22. [22]

    K-wav2vec 2.0: Joint decoding of graphemes and syllables,

    J. Kim and P. Kang, “K-wav2vec 2.0: Joint decoding of graphemes and syllables,” inProc. Interspeech, 2022

  23. [23]

    Cross-lingual and multilingual asr for scandinavian languages,

    R. ˇCerniavski, “Cross-lingual and multilingual asr for scandinavian languages,” Master’s thesis, Uppsala Univ., 2022

  24. [24]

    Multilingual asr for east african languages,

    M. Elamin et al., “Multilingual asr for east african languages,” inProc. AfricaNLP Workshop, 2023

  25. [25]

    Crossing language identification: Multilingual asr framework,

    O. H. Anidjar et al., “Crossing language identification: Multilingual asr framework,”Mach. Learn. Appl., vol. 13, p. 100489, 2024

  26. [26]

    Fine-tuning strategies for faster inference using speech self-supervised models: A comparative study,

    S. Zaiem, R. Algayres, T. Parcollet, S. Essid, and M. Ravanelli, “Fine-tuning strategies for faster inference using speech self-supervised models: A comparative study,” inProc. IEEE Int. Conf. Acoust., Speech, Signal Process. Workshops (ICASSPW), 2023, pp. 1–5

This paper was first reviewed by glm-5.2 on July 8, 2026.