Pith. sign in

REVIEW 4 major objections 6 minor 45 references

Improving Cross-Lingual Phonetic Representation of Low-Resource Languages Through Language Similarity Analysis

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Selecting just three phonologically similar source languages cuts phoneme error by 55.6% relative over monolingual training for low-resource languages.

desk verdict A practically useful source-selection study for low-resource phoneme recognition, with one load-bearing gap: the G2P transcriptions that drive both the similarity metric and the training/evaluation labels are never validated. read the letter →

arxiv 2501.06810 v1 pith:5GIMR2OZ submitted 2025-01-12 eess.AS cs.CLcs.SD

Minu Kim , Kangwook Jang , Hoirin Kim This is my paper · ORCID
classification eess.AScs.CLcs.SD
keywords sourcelanguageselectioncross-lingualphonemerecognitionlow-resourcespeechsimilarityG2PIPAtranscriptionmultilingualtrainingConformer
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

This paper tries to establish that how you choose source languages for cross-lingual speech training matters more than how much data you use. The authors propose a practical rule: compute the cosine similarity between unigram phoneme distributions of candidate source languages and the target language, using automatic grapheme-to-phoneme (G2P) conversion, then train on the top three most similar languages. Applied to 11 low-resource languages from three families, this rule improves average phoneme error rate by 55.6% relative over monolingual training and beats a large self-supervised model fine-tuned on each target. A sympathetic reader would care because it offers a cheap, data-free selection strategy for endangered and low-resource languages, and it explains conflicting earlier results about whether same-family training helps.

What carries the argument

The carrying object is the phoneme-distribution vector $p_A$ for each language, built by converting the training corpus text to IPA with a G2P model and counting unigram frequencies. Cosine similarity $\cos(p_A, p_B)$ between these vectors defines a similarity matrix, and the training set for a target is the top $k$ ($k=3$) most similar languages via Equation (3a). The same vectors drive PCA and KDE visualizations, and a parallel typology-based metric from Grambank features is used only as validation. The work that Equation (3a) does is to turn a vague notion of 'phonetic proximity' into a concrete, reproducible dataset-selection step that the experiments show predicts transfer success.

What would settle it

Take a held-out low-resource language, recompute the top-three source set using gold IPA transcripts or a different G2P tool instead of Epitran, and retrain the identical Conformer model; if the average phoneme error rate no longer beats monolingual training by a wide margin, or the selected languages change substantially, then the reported gain depends on the G2P tool rather than on genuine phonological similarity.

Watch

Extended reading notes

Core claim

The paper claims that for phoneme recognition in low-resource languages, a simple, parameter-free selection rule—take the three languages whose G2P-derived unigram phoneme distributions have the highest cosine similarity to the target—consistently outperforms monolingual training, family-based multilingual training, and training on all available languages. With the phoneme-level language model, the selected-language model reaches an average phoneme error rate of 25.3% across 11 low-resource languages, against 57.0% for monolingual training, a relative improvement of 55.6%; it also beats the fine-tuned XLSR-53 self-supervised model (32.2% average PER) despite using far less pre-training data and fewer parameters. Within a family, higher phonological similarity brings larger gains, but low-similarity family members can make family-based training worse than monolingual training, explaining prior contradictory findings.

Load-bearing premise

The whole selection rule assumes that cosine similarity between automatic G2P-derived unigram phoneme distributions faithfully measures how much one language's speech data helps train another; if the G2P conversions are noisy, two languages can look similar for the wrong reasons.

Editorial extensions

If this is right

  • If the claim holds, a low-resource language needs only a few hours of target speech plus text corpora of three well-chosen related languages to get phoneme recognition quality above a 300M-parameter self-supervised model.
  • Within-family multilingual training is not automatically beneficial; it helps for phonetically homogeneous families such as Turkic but can hurt for diverse families such as Indo-Iranian and Afro-Asiatic.
  • The same G2P-derived similarity matrix can be reused for any new target language without retraining, since the source set is chosen per target.
  • The phoneme-level language model amplifies the advantage of phonologically similar sources, so selection and decoding-side language modeling compound.
  • The method requires only text data for candidate source languages, making it applicable even when those sources have no speech recordings of their own.

Reading between the lines

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

  • The selection rule inherits any systematic bias of the G2P tools, since both similarity and training labels come from G2P output; using gold IPA annotations or a second G2P tool as a robustness check would show how much of the 55.6% gain is phonological versus tool-specific.
  • The fixed $k=3$ likely hides a curve: the optimal number of sources may vary with the target's data size and family homogeneity, so treating $k$ as a per-language hyperparameter is a natural extension.
  • Because the typology-based metric aligns with the corpus-based one for family similarity, a text-free version of this selection rule might be possible using Grambank features alone for languages without digitized text corpora.
  • The method implicitly assumes the Common Voice text is representative of each language's phoneme inventory; low-quality or non-phonetic orthographies would weaken the similarity signal.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a source-language selection method for cross-lingual phoneme recognition in low-resource languages. It computes cosine similarity between unigram phoneme distributions obtained by G2P conversion of text in the Common Voice corpus, selects the three most similar languages for each target (Eq. 3a), and trains a Conformer-based IPA transcription model on the target plus selected sources (Eq. 3b). Experiments across 11 low-resource languages from three families compare monolingual, family-based, all-languages, and similarity-based training, with and without a phoneme-level language model, and also compare against fine-tuned XLSR-53. The headline result is a 55.6% relative PER improvement over monolingual training and a 21.4% relative improvement over the fine-tuned SSL model, averaged over the 11 languages.

Significance. The core idea is practical and timely: rather than relying on expensive SSL pretraining or arbitrary source choices, the paper proposes a simple, externally defined phonological-similarity metric and tests it on 11 low-resource languages from three families. The main strengths are the consistency of the trend across most languages, the use of a metric that is not fitted to the PER results, and the explicit comparison with several alternative training-set configurations. If the label-side concerns are addressed, the method could be a useful baseline and a step toward principled source-language selection for low-resource ASR. The paper would be strengthened by making the experimental evidence more robust: reporting repeated runs with variance, adding a matched random-source control, and validating the G2P transcriptions that drive both selection and evaluation.

major comments (4)
  1. [Section II-B/II-C and Section III] The same G2P conversions are used for source selection (Eq. 1), training targets (Eq. 3b), and evaluation, yet no external validation of the G2P output is reported for any of the 22 languages. Because Epitran, CommonVoiceUtils, and Gruut may share systematic biases (e.g., non-standard symbol choices, diacritic loss), languages could appear phonologically similar because their G2P outputs are distorted in the same way. Since the headline claim is a 55.6% relative improvement, the PER numbers need to be demonstrably about linguistic phonology rather than label artifacts. I am not claiming the metric is circular, since it is external to model performance, but the label validity is load-bearing. Please report a G2P accuracy check against a small manually transcribed IPA set or an existing pronunciation dictionary for several languages, and/or compare the source sets selected by the corpus-based metric with those selected by typological or hand-built phoneme inventories.
  2. [Section IV-B, Table II] The paper lacks a random-selection control of matched training size. The "all" condition uses all 22 languages (more than 10 times the data, as shown in Fig. 5), and the "family" condition uses different numbers of source languages, so neither condition isolates the effect of similarity from the effect of adding more supervised data. To support the claim that phonological similarity, not merely extra data, drives the improvement, report a control that samples three source languages uniformly at random for each target, with repeated draws and a comparable total training duration to the corpus-sim condition. Without this baseline, the observed gains could largely reflect the benefit of adding any additional labeled speech.
  3. [Table II] Table II reports a single PER per condition, with no standard deviations, number of runs, or significance tests. Some comparisons are close (e.g., Hindi corpus-sim 22.6 vs. monolingual 23.2; Sakha 19.2 vs. 22.2), so the word "consistently" is not yet supported by the evidence. Please report at least three runs with different seeds and provide paired significance tests across the 11 low-resource languages, or otherwise quantify the variability of the reported averages.
  4. [Abstract and Section IV-B] The comparison with XLSR-53 is not matched: XLSR-53 is fine-tuned on the target low-resource language alone, whereas the Conformer corpus-sim model is trained on the target plus three source languages. The statement that the proposed approach "surpasses" a large SSL model therefore conflates the source-selection benefit with the use of additional labeled data. Please either fine-tune XLSR-53 on the same multilingual training set used by the corpus-sim model, or restrict the comparison to an equal-data setting, before claiming superiority over SSL.
minor comments (6)
  1. [Eq. (3a)] The notation "argmax" with |S|=3 is not literally an argmax; please specify that S contains the three languages with the largest cosine similarities to the target.
  2. [Eq. (3b)] The condition "∀ϕj ∈ ϕ, ϕj ∈ ∪_{Li∈S} P_language_i" is type-incorrect: ϕ is a sequence, while the union is a set of phonemes. Please clarify that each phoneme in the sequence must belong to the union, and state whether the target language's own phonemes are also included in the training set.
  3. [Section II-B, Eq. (2b)] The KDE contour level c=0.1 is a free choice; please state whether the qualitative conclusions in Section IV-A are sensitive to this value.
  4. [Section III] The reference [36] for the "CVC recipe" appears to be a semantic segmentation paper rather than the intended speech-recognition recipe; please verify and correct this citation.
  5. [Section IV-A] The phrase "significant improvements" in the family-based discussion is not backed by any statistical test; please rephrase or add formal significance testing.
  6. [Table II and Section IV-B] The 55.6% relative improvement is for the PLM-based rows (57.0 to 25.3); without the PLM the relative gain is approximately 52.8% (57.6 to 27.2). Please state this explicitly so the abstract claim is unambiguous about which configuration it refers to.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the source-language selection metric is an external linguistic measure, fixed a priori, and the reported PER gains are empirical rather than forced by construction.

full rationale

The paper's central claim is that training on the top-3 phonologically similar languages, selected by cosine similarity of G2P-derived unigram phoneme distributions (Eq. 1 and Eq. 3a), yields a 55.6% relative PER improvement over monolingual training. This selection rule is not fitted to the PER results: k=3 and the cosine metric are fixed choices made before training, and no equation in the paper defines the PER improvement as a function of the similarity score. The selection is therefore not a fitted input renamed as a prediction. There is no load-bearing self-citation chain: the cited Grambank, Epitran, CommonVoiceUtils, Gruut, Conformer, and XLSR-53 works are standard external tools or baselines, not prior results by the same authors invoked to forbid alternatives. The main caveat is a construct-validity concern rather than circularity: the same G2P pipeline converts text to IPA for both the similarity vectors and the training targets, and the paper reports no external IPA gold-standard validation for the 22 languages. Systematic G2P errors could make languages appear similar for spurious reasons, and the absence of a random-selection control means part of the gain could come from adding any extra source data. However, this is an empirical confounding risk, not a definitional equivalence: the paper never reduces the reported 55.6% improvement to the selection score by construction. The derivation chain is therefore self-contained against external benchmarks, and the circularity score is low.

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

All load-bearing choices are domain assumptions about what the phoneme-distribution similarity captures; no parameters are fitted to the reported PER values. The top-3 count and the contour level are hand-set constants. No new entities are postulated.

free parameters (4)
  • number of selected source languages (|S|) = 3
    Equation (3a) fixes the top-3 selection without an ablation study; performance could depend on this constant.
  • KDE contour level c = 0.1
    Equation (2b) sets the contour at density 0.1; used only for visualization, not for the recognition results.
  • low-resource threshold = 15 hours
    Defines which target languages are evaluated; adopted from prior work but arbitrary.
  • G2P post-adjustment rules = unspecified
    Manual decisions to omit voice quality symbols, remove stress diacritics, and merge phoneme pairs such as /S/ with /sj/; these change the phoneme distributions used for selection.
assumptions (4)
  • domain assumption Unigram phoneme distribution cosine similarity is a valid proxy for cross-lingual phonetic transfer
    Central to the source-selection rule in Eq. (3a); the paper does not test whether higher-order phonotactics or acoustic realization matter.
  • domain assumption G2P conversions are accurate for all 22 languages
    Both the phoneme-distribution vectors and the IPA training targets come from Epitran, CommonVoiceUtils, and Gruut; no validation of their output is reported.
  • domain assumption Common Voice validated utterances are reliable speech data
    The model training and evaluation rest on crowd-sourced validation without additional quality checks.
  • domain assumption Typological similarity from Grambank aligns with phonological similarity
    Used to corroborate family-level findings in Fig. 4, but not used to select training languages.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Cross-Lingual Phonetic Representation of Low-Resource Languages Through Language Similarity Analysis." pith.science (2026). https://pith.science/paper/5GIMR2OZ

@misc{pith2026250106810,
  author       = {Pith},
  title        = {Pith review of: Improving Cross-Lingual Phonetic Representation of Low-Resource Languages Through Language Similarity Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5GIMR2OZ}},
  note         = {Machine review of arXiv:2501.06810}
}
read the original abstract

This paper examines how linguistic similarity affects cross-lingual phonetic representation in speech processing for low-resource languages, emphasizing effective source language selection. Previous cross-lingual research has used various source languages to enhance performance for the target low-resource language without thorough consideration of selection. Our study stands out by providing an in-depth analysis of language selection, supported by a practical approach to assess phonetic proximity among multiple language families. We investigate how within-family similarity impacts performance in multilingual training, which aids in understanding language dynamics. We also evaluate the effect of using phonologically similar languages, regardless of family. For the phoneme recognition task, utilizing phonologically similar languages consistently achieves a relative improvement of 55.6% over monolingual training, even surpassing the performance of a large-scale self-supervised learning model. Multilingual training within the same language family demonstrates that higher phonological similarity enhances performance, while lower similarity results in degraded performance compared to monolingual training.

Figures

Figures reproduced from arXiv: 2501.06810 by the authors.

Figure 1
Figure 1. The Conformer-based model is trained to infer IPA sequences for [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The language similarity matrix measures cosine similarity between the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. (Left) Language family contours, based on phoneme distribution [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: The training data amounts for three representative languages from each [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 41 canonical work pages

  1. [1]

    Challenges and opportunities in processing low resource languages: A study on persian,

    M. Shamsfard, “Challenges and opportunities in processing low resource languages: A study on persian,” in Proc. International conference language technologies for all , 2019

    work page 2019

  2. [2]

    An investigation on ini- tialization schemes for multilayer perceptron training using multilingual data and their effect on asr performance,

    N. T. Vu, W. Breiter, F. Metze, and T. Schultz, “An investigation on ini- tialization schemes for multilayer perceptron training using multilingual data and their effect on asr performance,” in Proc. Interspeech, 2012, pp. 2586–2589

    work page 2012

  3. [3]

    Multilingual models with language embeddings for low- resource speech recognition,

    L.-M. Lam-Yee-Mui, W. B. Kheder, V .-B. Le, C. Barras, and J.-L. Gauvain, “Multilingual models with language embeddings for low- resource speech recognition,” in 2nd Annual Meeting of the ELRA/ISCA SIG on Under-resourced Languages , 2023, pp. 83–87

    work page 2023

  4. [4]

    Towards a common under- standing of contributing factors for cross-lingual transfer in multilingual language models: A review,

    F. Philippy, S. Guo, and S. Haddadan, “Towards a common under- standing of contributing factors for cross-lingual transfer in multilingual language models: A review,” in Proc. ACL, 2023, pp. 5877–5891

    work page 2023

  5. [5]

    Choosing transfer languages for cross- lingual learning,

    Y .-H. Lin, C.-Y . Chen, J. Lee, Z. Li, Y . Zhang, M. Xia, S. Rijhwani, J. He, Z. Zhang, X. Ma et al., “Choosing transfer languages for cross- lingual learning,” in Proc. ACL, 2019, pp. 3125–3135

    work page 2019

  6. [6]

    Sequence-based multi-lingual low resource speech recognition,

    S. Dalmia, R. Sanabria, F. Metze, and A. W. Black, “Sequence-based multi-lingual low resource speech recognition,” in Proc. ICASSP, 2018, pp. 4909–4913

    work page 2018

  7. [7]

    Exploiting adapters for cross-lingual low-resource speech recognition,

    W. Hou, H. Zhu, Y . Wang, J. Wang, T. Qin, R. Xu, and T. Shinozaki, “Exploiting adapters for cross-lingual low-resource speech recognition,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 30, pp. 317–329, 2021

    2021

  8. [8]

    Improving cross-lingual transfer learning for end-to-end speech recognition with speech translation,

    C. Wang, J. Pino, and J. Gu, “Improving cross-lingual transfer learning for end-to-end speech recognition with speech translation,” in Proc. Interspeech, 2020, pp. 4731–4735

    work page 2020

Show all 45 references
  1. [9]

    Cross-lingual subspace gaussian mixture models for low-resource speech recognition,

    L. Lu, A. Ghoshal, and S. Renals, “Cross-lingual subspace gaussian mixture models for low-resource speech recognition,” IEEE/ACM Trans- actions on Audio, Speech, and Language Processing , vol. 22, no. 1, pp. 17–27, 2013

    2013

  2. [10]

    Cross-language end-to-end speech recognition research based on transfer learning for the low-resource tujia language,

    C. Yu, Y . Chen, Y . Li, M. Kang, S. Xu, and X. Liu, “Cross-language end-to-end speech recognition research based on transfer learning for the low-resource tujia language,” Symmetry, vol. 11, no. 2, p. 179, 2019

    2019

  3. [11]

    Linguistic disparities in cross-language automatic speech recognition transfer from arabic to tashlhiyt,

    G. Zellou and M. Lahrouchi, “Linguistic disparities in cross-language automatic speech recognition transfer from arabic to tashlhiyt,” Scientific Reports, vol. 14, no. 1, p. 313, 2024

    2024

  4. [12]

    Cross-lingual portability of chinese and english neural network features for french and german lvcsr,

    C. Plahl, R. Schlüter, and H. Ney, “Cross-lingual portability of chinese and english neural network features for french and german lvcsr,” in Proc. ASRU, 2011, pp. 371–376

    2011

  5. [13]

    Development of multi- lingual asr using globalphone for less-resourced languages: The case of ethiopian languages

    M. Y . Tachbelie, S. T. Abate, and T. Schultz, “Development of multi- lingual asr using globalphone for less-resourced languages: The case of ethiopian languages.” in Proc. Interspeech, 2020, pp. 1032–1036

    2020

  6. [14]

    Using large self-supervised models for low-resource speech recognition

    D. Krishna, P. Wang, and B. Bozza, “Using large self-supervised models for low-resource speech recognition.” in Proc. Interspeech , 2021, pp. 2436–2440

    2021

  7. [15]

    Cam: A cross-lingual adaptation framework for low-resource language speech recognition,

    Q. Hu, Y . Zhang, X. Zhang, Z. Han, and X. Yu, “Cam: A cross-lingual adaptation framework for low-resource language speech recognition,” Information Fusion, p. 102506, 2024

    2024

  8. [16]

    Cross-language knowledge transfer using multilingual deep neural network with shared hidden layers,

    J.-T. Huang, J. Li, D. Yu, L. Deng, and Y . Gong, “Cross-language knowledge transfer using multilingual deep neural network with shared hidden layers,” in Proc. ICASSP, 2013, pp. 7304–7308

    2013

  9. [17]

    Multilingual speech recognition for turkic languages,

    S. Mussakhojayeva, K. Dauletbek, R. Yeshpanov, and H. A. Varol, “Multilingual speech recognition for turkic languages,” Information, vol. 14, no. 2, p. 74, 2023

    2023

  10. [18]

    A comparison between morphological complexity measures: typological data vs. lan- guage corpora,

    C. Bentz, T. Ruzsics, A. Koplenig, and T. Samardzic, “A comparison between morphological complexity measures: typological data vs. lan- guage corpora,” in Proc. Workshop on computational linguistics for linguistic complexity, 2016, pp. 142–153

    2016

  11. [19]

    Frequency of occurrence of phonemes in conversational english,

    M. A. Mines, B. F. Hanson, and J. E. Shoup, “Frequency of occurrence of phonemes in conversational english,” Language and speech , vol. 21, no. 3, pp. 221–241, 1978

    1978

  12. [20]

    Grambank reveals the importance of genealogical constraints on linguistic diversity and highlights the impact of language loss,

    H. Skirgård, H. J. Haynie, D. E. Blasi, H. Hammarström, J. Collins, J. J. Latarche, J. Lesage, T. Weber, A. Witzlack-Makarevich, S. Passmore et al., “Grambank reveals the importance of genealogical constraints on linguistic diversity and highlights the impact of language loss,...

    2023

  13. [21]

    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 et al., “Conformer: Convolution-augmented transformer for speech recognition,” in Proc. Interspeech , 2020, pp. 5036–5040

    2020

  14. [22]

    Un- supervised cross-lingual representation learning for speech recognition,

    A. Conneau, A. Baevski, R. Collobert, A. Mohamed, and M. Auli, “Un- supervised cross-lingual representation learning for speech recognition,” in Proc. Interspeech, 2021, pp. 2426–2430

    2021

  15. [23]

    Documenting and researching endangered languages: The pangloss collection,

    B. Michailovsky, M. Mazaudon, A. Michaud, S. Guillaume, A. François, and E. Adamou, “Documenting and researching endangered languages: The pangloss collection,” LaNguagE documENtatIoN & coNSErvatIoN , vol. 8, 2014

    2014

  16. [24]

    Methods of language docu- mentation in the dobes program,

    P. Wittenburg, U. Mosel, and A. Dwyer, “Methods of language docu- mentation in the dobes program,” in Proc. LREC. Workshop on Tools and Resources in Field Lingusitics , 2002, pp. 36–42

    2002

  17. [25]

    Simple and effective zero-shot cross- lingual phoneme recognition,

    Q. Xu, A. Baevski, and M. Auli, “Simple and effective zero-shot cross- lingual phoneme recognition,” in Proc. Interspeech , 2022, pp. 2113– 2117

    2022

  18. [26]

    Dif- ferentiable allophone graphs for language-universal speech recognition,

    B. Yan, S. Dalmia, D. R. Mortensen, F. Metze, and S. Watanabe, “Dif- ferentiable allophone graphs for language-universal speech recognition,” in Proc. Interspeech, 2021, pp. 2471–2475

    2021

  19. [27]

    Universal automatic phonetic transcription into the international phonetic alphabet,

    C. Taguchi, Y . Sakai, P. Haghani, and D. Chiang, “Universal automatic phonetic transcription into the international phonetic alphabet,” in Proc. Interspeech, 2023, pp. 2548–2552

    2023

  20. [28]

    Phono- logically aware neural model for named entity recognition in low resource transfer settings,

    A. Bharadwaj, D. R. Mortensen, C. Dyer, and J. G. Carbonell, “Phono- logically aware neural model for named entity recognition in low resource transfer settings,” in Proc. EMNLP, 2016, pp. 1462–1472

    2016

  21. [29]

    Kernel density estimation and its application,

    S. W˛ eglarczyk, “Kernel density estimation and its application,” in ITM web of conferences , vol. 23. EDP Sciences, 2018, p. 00037

    2018

  22. [30]

    Improving the bandwidth selection in kernel equating,

    B. Andersson and A. A. von Davier, “Improving the bandwidth selection in kernel equating,” Journal of Educational Measurement, vol. 51, no. 3, pp. 223–238, 2014

    2014

  23. [31]

    Haspelmath, M

    M. Haspelmath, M. S. Dryer, D. Gil, and B. Comrie, The world atlas of language structures online. Max Planck Digital Library Munich, 2008

    2008

  24. [32]

    Common 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, “Common voice: A massively-multilingual speech corpus,” in Proc. LREC, 2020

    2020

  25. [33]

    Epitran: Precision g2p for many languages,

    D. R. Mortensen, S. Dalmia, and P. Littell, “Epitran: Precision g2p for many languages,” in Proc. LREC, 2018

    2018

  26. [34]

    Predicting positive transfer for improved low-resource speech recognition using acoustic pseudo-tokens,

    N. San, G. Paraskevopoulos, A. Arora, X. He, P. Kaur, O. Adams, and D. Jurafsky, “Predicting positive transfer for improved low-resource speech recognition using acoustic pseudo-tokens,” arXiv preprint arXiv:2402.02302, 2024

    2024 arXiv

  27. [35]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. NeurIPS, vol. 30, 2017

    2017

  28. [36]

    Espnet: Efficient spatial pyramid of dilated convolutions for semantic segmentation,

    S. Mehta, M. Rastegari, A. Caspi, L. Shapiro, and H. Hajishirzi, “Espnet: Efficient spatial pyramid of dilated convolutions for semantic segmentation,” in Proceedings of the european conference on computer vision (ECCV), 2018, pp. 552–568

    2018

  29. [37]

    Phoneme level language models for sequence based low resource asr,

    S. Dalmia, X. Li, A. W. Black, and F. Metze, “Phoneme level language models for sequence based low resource asr,” in ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Pro- cessing (ICASSP). IEEE, 2019, pp. 6091–6095

    2019

  30. [38]

    Roman jakobson’s conception of «sprachbund»,

    H. W. Schaller, “Roman jakobson’s conception of «sprachbund»,” Cahiers du Centre de Linguistique et des Sciences du Langage , no. 9, pp. 207–212, 1997

    1997

  31. [39]

    The history of turkic,

    L. Johanson, “The history of turkic,” in The Turkic Languages. Rout- ledge, 2021, pp. 83–120

    2021

  32. [40]

    On the kurdish language,

    P. G. Kreyenbroek, “On the kurdish language,” in The Kurds. Routledge, 2005, pp. 62–73

    2005

  33. [41]

    The performance evaluation of attention-based neural asr under mixed speech input,

    B. He and M. Radfar, “The performance evaluation of attention-based neural asr under mixed speech input,” arXiv preprint arXiv:2108.01245, 2021

    2021 arXiv

  34. [42]

    Improving automatic speech recognition performance for low-resource languages with self-supervised models,

    J. Zhao and W.-Q. Zhang, “Improving automatic speech recognition performance for low-resource languages with self-supervised models,” IEEE Journal of Selected Topics in Signal Processing , vol. 16, no. 6, pp. 1227–1241, 2022

    2022

  35. [43]

    Multilingual self- supervised speech representations improve the speech recognition of low-resource african languages with codeswitching,

    T. Ogunremi, C. D. Manning, and D. Jurafsky, “Multilingual self- supervised speech representations improve the speech recognition of low-resource african languages with codeswitching,” in Proc. Workshop on Computational Approaches to Linguistic Code-Switching , 2023, pp. 83–88

    2023

  36. [44]

    Meta-adaptable- adapter: Efficient adaptation of self-supervised models for low-resource speech recognition,

    Y . Chen, H. Zhang, X. Yang, W. Zhang, and D. Qu, “Meta-adaptable- adapter: Efficient adaptation of self-supervised models for low-resource speech recognition,” Neurocomputing, p. 128493, 2024

    2024

  37. [45]

    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,” in Proc. NeurIPS, vol. 33, 2020, pp. 12 449–12 460

    2020

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.