Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Towards a Unified Benchmark for Arabic Pronunciation Assessment: Quranic Recitation as Case Study

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper introduces QuranMB.v1, the first publicly available benchmark for mispronunciation detection in Modern Standard Arabic using Qur'anic recitation, and reports that the strongest baseline—a multilingual HuBERT model trained on…

desk verdict Useful first benchmark for Arabic MDD, but its test set is built from the same confusion matrix as the training data, so the reported F1 measures an in-distribution error model rather than real-world generalization. read the letter →

arxiv 2506.07722 v2 pith:V5MTD3C7 submitted 2025-06-09 cs.SD eess.AS

Yassine El Kheir , Omnia Ibrahim , Amit Meghanani , Nada Almarwani , Hawau Olamide Toyin , Sadeen Alharbi , Modar Alfadly , Lamya Alkanhal
show 7 more authors
Ibrahim Selim Shehab Elbatal Salima Mdhaffar Thomas Hain Yasser Hifny Mostafa Shahin Ahmed Ali
This is my paper
classification cs.SDeess.AS
keywords ArabicpronunciationassessmentmispronunciationdetectionQuranicrecitationModernStandardsyntheticspeechaugmentationconfusionmatrixself-supervisedmodelsbenchmarkdataset
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The authors set out to give Arabic pronunciation-assessment research a common yardstick by releasing QuranMB.v1, a public test benchmark built from 98 Qur'anic verses read by 18 native Arabic speakers with deliberately injected pronunciation errors. They accompany it with a full pipeline: a 68-phoneme inventory tailored to Modern Standard Arabic, an in-the-wild real-speech training set, and 52 hours of synthetic speech in which errors are inserted by a confusion matrix and rendered by text-to-speech. Using frozen self-supervised speech encoders with CTC decoding, they report that the best baseline, mHuBERT trained on real plus synthetic data, reaches 29.88% F1 on mispronunciation detection. The headline point is not that the problem is solved but that a standardized resource now exists on which future approaches can be compared, and the low score quantifies how much room remains.

What carries the argument

The load-bearing object is QuranMB.v1, the test benchmark itself, together with the error-generation protocol used to build it: a confusion matrix derived from phoneme-similarity data maps each Arabic character or diacritic to likely mispronounced counterparts, including deletions. The same matrix is used to corrupt canonical transcripts before TTS rendering, yielding the 52-hour synthetic training corpus; this symmetry is what makes the controlled errors fully annotated. The recognition pipeline is a frozen self-supervised encoder whose layer-weighted features feed a two-layer Bi-LSTM with CTC loss, with greedy decoding producing phoneme sequences that are compared against phonetizer-derived targets. The evaluation follows the standard TA/TR/FA/FR and CD/ED categorization, with F1 on the reject class as the headline metric.

What would settle it

Record a second test set in which Arabic speakers read the same verses freely without being told which errors to produce, have experts annotate the resulting mispronunciations, and run the same best model on it. If its F1 drops far below 29.88% or the distribution of error pairs diverges from the confusion matrix, the benchmark's claim to represent realistic MSA mispronunciations would be falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that Qur'anic recitation, read in Modern Standard Arabic without tajweed constraints, is a workable case study for benchmarking Arabic mispronunciation detection, and that a publicly released test set plus a reproducible training pipeline can support fair comparisons. The authors assert that their synthetic TTS corpus, generated by randomly modifying four characters or diacritics per canonical transcript according to a confusion matrix derived from phoneme similarity data, is competitive with wild-collected speech for training MDD models: mHuBERT trained on the combined corpus attains the best F1 (29.88%), and even TTS-only training beats several English-only SSL baselines trained on real speech. They also claim the 68-phoneme inventory, which merges emphatic-context vowel variants into single phonemes and marks gemination by doubled symbols, is appropriate for the task.

Load-bearing premise

The whole evaluation rests on a table of which Arabic sounds are commonly swapped for which; if that table does not reflect how real speakers actually mispronounce Modern Standard Arabic, both the synthetic training data and the scripted test errors are unrealistic.

Editorial extensions

If this is right

  • Future Arabic mispronunciation detection systems can be compared on a stable public test set instead of private, ad hoc evaluations.
  • Because synthetic TTS-only training already rivals real-speech-only training for some baselines, controlled synthetic mispronunciation data is a credible route around annotation scarcity.
  • Multilingual self-supervised pretraining transfers to Arabic better than English-only pretraining in these experiments, so subsequent CAPT work should start from multilingual encoders.
  • The reported F1 below 30% means current models falsely accept most mispronunciations, so data curation and specialized modeling, not just larger pretraining, are the near-term levers.
  • The 68-phoneme inventory and the evaluation protocol become reusable infrastructure for other Arabic pronunciation tasks, including future extensions to Tajweed-oriented checking.

Reading between the lines

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

  • Because the same confusion matrix appears to generate both the synthetic training errors and the scripted test errors, the reported F1 may be optimistic relative to natural, unscripted mispronunciations; a benchmark containing spontaneously occurring errors would be needed to test that transfer.
  • The benchmark deliberately reads MSA without Tajweed rules, so it measures segmental errors such as consonant substitutions rather than prosodic recitation mistakes; researchers applying it to recitation quality should not expect it to cover those.
  • The protocol of cueing speakers with highlighted modified text likely produces non-spontaneous, carefully timed errors, which could make the test either easier or harder than natural errors in ways the current numbers do not reveal.
  • Extending the same confusion-matrix and TTS pipeline to second-language learners would require enlarging the phoneme inventory with non-Arabic sounds, as the authors themselves note for future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces QuranMB.v1, a benchmark test set for mispronunciation detection and diagnosis (MDD) in Modern Standard Arabic using Qur'anic recitation. It consists of 98 verses read by 18 native Arabic speakers who were instructed to produce specific pronunciation errors selected from a confusion matrix. The authors also build a 52-hour synthetic training corpus by modifying canonical vowelized transcripts with simulated errors and synthesizing speech with seven TTS voices. Several SSL-based baseline models (Wav2vec2, HuBERT, WavLM, mHuBERT) are evaluated under three training configurations, with the best reported result being an F1 of 29.88% for mHuBERT trained on combined real and synthetic data. The paper claims this is the first publicly available benchmark for Arabic mispronunciation detection in the Qur'anic recitation setting.

Significance. The contribution is timely and potentially useful: public MDD resources for Arabic are scarce, and the paper provides a documented pipeline spanning a specialized phoneme set, TTS-based error augmentation, a test set, and baseline evaluations. The release of QuranMB.v1, if actually public and properly validated, would give the community a controlled starting point for Arabic MDD research. However, the benchmark's external validity is currently unestablished because the test errors are generated from the same confusion matrix used to synthesize training data, and because no natural learner errors or expert annotations are involved. The baseline performance numbers should therefore be interpreted as measuring a model's fit to the authors' simulated error distribution, not its ability to detect natural Arabic mispronunciations.

major comments (3)
  1. [Sections 2.2 and 2.4] Both the synthetic training errors (Section 2.2) and the test-set ground truth (Section 2.4) are generated by selecting characters or diacritics and replacing them using the same confusion matrix derived from the authors' prior work [17]. The test errors are additionally acted by native speakers instructed to produce those specific substitutions, so the test set does not contain independently observed learner mispronunciations. Consequently, the F1 scores in Table 2 (e.g., 29.88 for mHuBERT on CMV-Ar+TTS) measure agreement with the authors' simulated error model, not performance on naturally occurring errors. The manuscript should either validate the confusion matrix against expert-annotated real learner errors, use a held-out set of natural errors for testing, or explicitly limit the benchmark's claims to controlled synthetic-error scenarios.
  2. [Section 2.3] The test set contains no naturally occurring mispronunciations: all errors are deliberately produced by native Arabic speakers following displayed instructions. There is no reported quality check of whether the speakers actually produced the intended substitutions, no inter-speaker consistency analysis, and no expert phonetician annotation of the recordings. This makes it difficult to know whether the ground-truth labels correspond to the acoustics in the audio. Please add quality-control statistics, such as agreement between intended and perceived errors or re-annotation of a subset by experts, and describe how the recorded speech was verified against the intended error patterns.
  3. [Sections 2.1, 2.2, and 2.4] The abstract and contributions claim that QuranMB.v1 is the 'first publicly available test set' and state that all models and datasets are available at the Hugging Face link, but the body repeatedly uses the future tense: the CMV-Ar corpus 'will be made publicly available' (Section 2.1), the TTS dataset 'will be publicly available' (Section 2.2), and the confusion dictionary 'will be publicly available' (Section 2.4). This inconsistency makes the central release claim unverifiable. Please provide stable identifiers (e.g., dataset card, DOI) and state the current accessibility, license, and access terms for QuranMB.v1 and the training corpora.
minor comments (5)
  1. [Section 2.2] The sentence 'we randomly select four characters and/or diacritics and modify them based on a predefined confusion pairs matrix' is underspecified; please describe the exact sampling procedure, whether the number of modified tokens is fixed per transcript, and how the confusion matrix probabilities are applied, as this is essential for reproducibility.
  2. [Section 3.4, Eq. (1)] The definitions of Precision and Recall are nonstandard, and the roles of 'True Reject' and 'False Reject' should be clarified in the text. A brief justification that these quantities are computed only over the error-detection decision would help readers interpret Table 2.
  3. [Table 2] The table caption contains the misspelling 'mispronunications.' Additionally, the columns TA, FR, FA, CD, and ED are used in the caption without definitions; please define all abbreviations in the caption or in the text.
  4. [Section 2.5] The paper mentions that Arabic has 34 phonemes in the Introduction but then defines a phoneme inventory of 68 phonemes. Please clarify how the 68-phoneme set relates to the 34-phoneme description, particularly whether it includes geminated variants, emphatic variants, and positional allophones.
  5. [Section 3.1] The SSL model variants include English-only and multilingual models, but no Arabic-specific pre-trained model. If no suitable Arabic SSL model exists, state this explicitly; otherwise, its absence from the baselines should be justified.

Circularity Check

2 steps flagged · score 6.0 of 10

QuranMB.v1 test labels are generated with the same confusion matrix used to synthesize the TTS training data, so the reported F1 measures self-consistency with the authors' error model rather than natural mispronunciation.

  1. self definitional [Sections 2.2, 2.3, and 2.4]
    "Given a canonical transcript, we randomly select four characters and/or diacritics and modify them based on a predefined confusion pairs matrix. The construction of this matrix is specified in Section 2.4. The confusion matrix is derived from phoneme similarity data extracted from [17]. The speakers were instructed to read the text in MSA at their normal tempo, disregarding Qur'anic tajweed rules, while deliberately producing the specified pronunciation errors."

    The same confusion matrix (from [17]) is used in Section 2.2 to generate synthetic training errors and, through the recording instructions in Section 2.3, to define the gold errors in QuranMB.v1. Therefore the Table 2 F1 score measures how well a model trained on that matrix's error distribution detects that same matrix's errors; it is not an estimate of performance on naturally occurring learner mispronunciations. The test set's ground truth is defined by the very error model used to train the model, making the reported benchmark performance self-referential by construction.

  2. self citation load bearing [Section 2.4, reference [17]]
    "The confusion matrix is derived from phoneme similarity data extracted from [17], ensuring that substitutions align with natural mispronunciation tendencies."

    [17] is the authors' own SpeechBlender paper, and it is the only justification offered for the claim that the substitutions align with natural mispronunciation tendencies. No independent, expert-annotated corpus of real L2 Arabic mispronunciations is used to validate this matrix within the present paper. The conclusion's statement that future work requires 'collection of diverse datasets, including L2 speakers' concedes that the current test set is not grounded in natural learner errors. Thus the central premise that QuranMB.v1 represents realistic MSA mispronunciations rests on an unvalidated self-citation.

full rationale

The paper contributes a real recorded speech corpus, a phoneme set, and an evaluation pipeline, and the low F1 scores are honestly reported; the work is not wholly circular. However, the evaluation loop is closed: the gold labels for QuranMB.v1 are constructed by applying the same confusion matrix, sourced from the authors' prior work [17], that also generates the synthetic TTS training data. Consequently, the headline F1=29.88 reflects agreement with the authors' simulated error model, and the benchmark's external validity for natural MSA mispronunciations is untested. The future-tense statements about release also mean the 'publicly available' status is not yet verifiable, though that is a verifiability issue rather than circularity. Because the central claim of a usable benchmark depends on this unvalidated, self-sourced error model, a moderate circularity score is warranted.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical or theoretical entities are introduced. The main constructs, QuranMB.v1 and the 68-phoneme set, are datasets and annotations, not invented entities in the schema's sense. The free parameters and axioms listed above are the load-bearing choices that shape the benchmark and its results.

free parameters (3)
  • modified characters per transcript = 4
    Section 2.2: 'we randomly select four characters and/or diacritics and modify them based on a predefined confusion pairs matrix.' This sets the error density in synthetic data without calibration to real mispronunciation rates.
  • confusion matrix entries = not reported (derived from [17])
    Section 2.4: the confusion matrix maps each character/diacritic to confusable counterparts based on phoneme similarity extracted from [17]. The specific probabilities and choices are not given, and both synthetic training and test errors depend on them.
  • number of TTS voices = 7 (5 male, 2 female)
    Section 2.2: seven single-speaker TTS systems are used; the voice count and gender proportion are design choices not justified against speaker diversity or domain coverage.
assumptions (4)
  • domain assumption Common Voice MSA speech aligns with linguistically driven transcript vowelization after in-house vowelization
    Section 2.1 and 2.2: the training corpus assumes that speech perfectly matches the vowelized transcriptions and that no character-level mispronunciations occur.
  • ad hoc to paper The confusion matrix derived from [17] reflects realistic mispronunciation patterns for MSA/Quranic recitation
    Section 2.4: the benchmark's ground truth errors are generated from this matrix; its validity is assumed without comparison to real learner errors, and it originates from the authors' own prior work.
  • domain assumption Halabi's phonetizer produces correct phoneme sequences for vowelized MSA
    Section 2.5: the phonetizer developed for speech synthesis is used as the ground truth for phoneme sequences in all datasets.
  • domain assumption Reading MSA without tajweed while producing targeted errors is a meaningful proxy for Quranic recitation MDD
    Section 2.3: speakers are instructed to disregard tajweed rules and deliberately produce specified errors; the benchmark does not include natural tajweed-governed recitation or natural learner errors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards a Unified Benchmark for Arabic Pronunciation Assessment: Quranic Recitation as Case Study." pith.science (2026). https://pith.science/paper/V5MTD3C7

@misc{pith2026250607722,
  author       = {Pith},
  title        = {Pith review of: Towards a Unified Benchmark for Arabic Pronunciation Assessment: Quranic Recitation as Case Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V5MTD3C7}},
  note         = {Machine review of arXiv:2506.07722}
}
read the original abstract

We present a unified benchmark for mispronunciation detection in Modern Standard Arabic (MSA) using Qur'anic recitation as a case study. Our approach lays the groundwork for advancing Arabic pronunciation assessment by providing a comprehensive pipeline that spans data processing, the development of a specialized phoneme set tailored to the nuances of MSA pronunciation, and the creation of the first publicly available test set for this task, which we term as the Qur'anic Mispronunciation Benchmark (QuranMB.v1). Furthermore, we evaluate several baseline models to provide initial performance insights, thereby highlighting both the promise and the challenges inherent in assessing MSA pronunciation. By establishing this standardized framework, we aim to foster further research and development in pronunciation assessment in Arabic language technology and related applications.

Figures

Figures reproduced from arXiv: 2506.07722 by the authors.

Figure 1
Figure 1. TTS Augmentation Pipeline Second, we generated a synthetic corpus using Text-to￾speech (TTS) models trained on fully vowelized text, en￾suring precise speech-to-phoneme alignment. The test set QuranMB.v1 consists of natural speech data that is collected with vowelization, following our detailed collec￾tion guidelines presented later in this section. 2.1. Training Set: CMV-Ar Our training corpus, CMV-Ar, incorporates… view at source ↗
Figure 3
Figure 3. Mispronunciation Detection Modeling Pipeline 2.5. Data Processing (Describe Phoneme Set) Since the goal of this work is to detect mispronounced sounds in Qur’an recitation, the model is designed to rec￾ognize the sequence of pronounced phonemes and com￾pare it to the target phoneme sequence. To accomplish this, correct phoneme transcriptions of the dataset are required as training targets for the model. As our fo￾cu… view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Automatic Pronunciation Error Detection and Correction of the Holy Quran's Learners Using Deep Learning

    eess.AS 2025-08 conditional novelty 6.0 of 10

    The authors release a rule-based Quran Phonetic Script, an 890-hour expert recitation dataset, and a multi-head CTC model that achieves 0.16% average phoneme error rate on held-out reciters.

Reference graph

Works this paper leans on

34 extracted references · 29 canonical work pages · cited by 1 Pith paper

  1. [17]

    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 Advances in Neural Infor- mation 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

    work page 2020

  2. [1]

    CAPT systems leverage advances in speech technology, curricu- lum management, and learner assessment to guide learn- ers toward improved pronunciation

    Introduction Computer-aided Pronunciation Training (CAPT) has be- come an essential tool for self-directed language learners by providing real-time feedback through systematic eval- uation and correction of pronunciation errors [1]. CAPT systems leverage advances in speech technology, curricu- lum management, and learner assessment to guide learn- ers tow...

    work page

  3. [2]

    Dataset Curation Due to the limited availability of phonetically labeled Arabic speech, we employed two complementary strate- gies for our training corpus. First, we collected an in- the-wild dataset of non-dialectal Arabic speech, assuming that speakers follow standardized vowelization rules, and automatically vowelized the corresponding transcription. a...

    arXiv 2025

  4. [3]

    We have used simi- lar setup as described in the SUPERB [19] to train the model

    Methodology Our framework for MDD leverages self-supervised learn- ing (SSL)-based speech models with subsequent temporal modeling, as depicted in Figure 3. We have used simi- lar setup as described in the SUPERB [19] to train the model. In SUPERB, the SSL models weights are frozen. 2https://huggingface.co/spaces/IqraEval/ArabicPhoneme Table 1: Data confi...

    work page

  5. [4]

    Key trends emerge in mul- tilingual capability, synthetic data utility, and task complexity

    Results and Analysis Table ?? summarizes the performance of SSL- based MDD models across dataset configurations on QuranMB.v1 test set. Key trends emerge in mul- tilingual capability, synthetic data utility, and task complexity. 4.1. Multilingual vs Monolingual SSL models The multilingual mHuBERT model consistently outper- forms monolingual SSL variants a...

    work page

  6. [5]

    Conclusion In this work, we have introduced the first comprehen- sive public benchmark for MSA pronunciation detection, providing detailed documentation of our data curation methodology, specialized phoneme set, and data augmen- tation approaches. The release of QuranMB.v1, our test dataset, represents a significant contribution as the first publicly avai...

    work page

  7. [6]

    Acknowledgment We thank the Saudi Data and Artificial Intelligence Au- thority (SDAIA) for hosting the Winter School where this work took place, and for the generous computing resources provided

    work page

  8. [7]

    The good- ness of pronunciation algorithm: a detailed performance study,

    S. Kanters, C. Cucchiarini, and H. Strik, “The good- ness of pronunciation algorithm: a detailed performance study,” 2009

    work page 2009

Show all 34 references
  1. [8]

    Multitask learning for arabic dialects identification and machine translation,

    M. Dhleima, M. C. Tourad, C. A. A. Telmoud, A. Ab- delmounaim, and M. F. Nanne, “Multitask learning for arabic dialects identification and machine translation,” in International Conference on Artificial Intelligence and its Applications in the Age of Digital Transformation . S...

    2024

  2. [9]

    The ef- fectiveness of computer assisted pronunciation training for foreign language learning by children,

    A. Neri, O. Mich, M. Gerosa, and D. Giuliani, “The ef- fectiveness of computer assisted pronunciation training for foreign language learning by children,” Computer As- sisted Language Learning , vol. 21, no. 5, pp. 393–408, 2008

    2008

  3. [10]

    Beyond orthography: Automatic recovery of short vowels and dialectal sounds in arabic,

    Y. E. Kheir, H. Mubarak, A. Ali, and S. A. Chowd- hury, “Beyond orthography: Automatic recovery of short vowels and dialectal sounds in arabic,” arXiv preprint arXiv:2408.02430, 2024

    2024 arXiv

  4. [11]

    Phonetic rules in arabic script,

    M. M. Al-Ghamdi, H. Al-Muhtasib, and M. Elshafei, “Phonetic rules in arabic script,” Journal of King Saud University-Computer and Information Sciences , vol. 16, pp. 85–115, 2004

    2004

  5. [12]

    Improv- ing mispronunciation detection and diagnosis for non- native learners of the arabic language,

    N. Alrashoudi, H. Al-Khalifa, and Y. Alotaibi, “Improv- ing mispronunciation detection and diagnosis for non- native learners of the arabic language,” Discover Com- puting, vol. 28, no. 1, p. 1, 2025

    2025

  6. [13]

    Automatic speech recognition (asr) systems for learning arabic lan- guage and al-quran recitation: a review,

    N. O. Balula, M. Rashwan, and S. Abdou, “Automatic speech recognition (asr) systems for learning arabic lan- guage and al-quran recitation: a review,” International Journal of Computer Science and Mobile Computing , vol. 10, no. 7, pp. 91–100, 2021

    2021

  7. [14]

    Phone-level pronunciation scoring and assessment for interactive language learning,

    S. M. Witt and S. J. Young, “Phone-level pronunciation scoring and assessment for interactive language learning,” Speech communication, vol. 30, no. 2-3, pp. 95–108, 2000

    2000

  8. [15]

    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,” arXiv preprint arXiv:1912.06670 , 2019

    1912 arXiv

  9. [16]

    Computer-assisted pronunciation train- ing—speech synthesis is almost all you need,

    D. Korzekwa, J. Lorenzo-Trueba, T. Drugman, and B. Kostek, “Computer-assisted pronunciation train- ing—speech synthesis is almost all you need,” Speech Communication, vol. 142, pp. 22–33, 2022

    2022

  10. [18]

    This phonetizer was optimized for phonetic coverage in speech synthesis

    developed for vowelized MSA. This phonetizer was optimized for phonetic coverage in speech synthesis. It employs a greedy algorithm to minimize the size of the speech corpus while maintaining comprehensive phonetic and prosodic coverage. The phonetic vocabulary includes diphon...

  11. [19]

    The mgb-2 challenge: Arabic multi- dialect broadcast media recognition,

    A. Ali, P. Bell, J. Glass, Y. Messaoui, H. Mubarak, S. Re- nals, and Y. Zhang, “The mgb-2 challenge: Arabic multi- dialect broadcast media recognition,” in 2016 IEEE Spo- ken Language Technology Workshop (SLT). IEEE, 2016, pp. 279–284

    2016

  12. [20]

    To obtain the phoneme sequence during inference, CTC greedy decoding is used

    loss on phoneme sequence. To obtain the phoneme sequence during inference, CTC greedy decoding is used. 3.1. SSL Model Variants We investigate both monolingual and multilingual SSL variants. Our base experiments use 94M-parameter models: English-only Wav2vec2 [9], HuBERT [21],...

  13. [21]

    A novel framework for mispronunciation detection of ara- bic phonemes using audio-oriented transformer models,

    S ¸. S. C ¸ alık, A. K¨ u¸ c¨ ukmanisa, and Z. H. Kilimci, “A novel framework for mispronunciation detection of ara- bic phonemes using audio-oriented transformer models,” Applied Acoustics, vol. 215, p. 109711, 2024

    2024

  14. [22]

    Empirical study on mis- pronunciation detection for tajweed rules during quran recitation,

    Y. S. Alsahafi and M. Asad, “Empirical study on mis- pronunciation detection for tajweed rules during quran recitation,” in 2024 6th ICCI . IEEE, 2024, pp. 39–45

    2024

  15. [23]

    Development of automated tajweed checking system for children in learning quran,

    M. A. Rahman, I. A. A. Kassim, T. A. Rahman, and S. Z. M. Muji, “Development of automated tajweed checking system for children in learning quran,” Evo- lution in Electrical and Electronic Engineering , vol. 2, no. 1, pp. 165–176, 2021

    2021

  16. [24]

    Quran recitation recog- nition using end-to-end deep learning,

    A. A. Harere and K. A. Jallad, “Quran recitation recog- nition using end-to-end deep learning,” arXiv preprint arXiv:2305.07034, 2023

    2023 arXiv

  17. [25]

    Speechblender: Speech augmentation framework for mispronunciation data generation,

    Y. E. Kheir, S. A. Chowdhury, A. Ali, H. Mubarak, and S. Afzal, “Speechblender: Speech augmentation framework for mispronunciation data generation,” arXiv preprint arXiv:2211.00923, 2022

    2022 arXiv

  18. [26]

    Phonetic inventory for an arabic speech corpus,

    N. Halabi and M. Wald, “Phonetic inventory for an arabic speech corpus,” in LREC’16, 2016, pp. 734–738

    2016

  19. [27]

    Superb: Speech processing universal perfor- mance benchmark,

    S. wen Yang, P.-H. Chi, Y.-S. Chuang, C.-I. J. Lai, and K. L. et al., “Superb: Speech processing universal perfor- mance benchmark,” in Interspeech 2021, 2021, pp. 1194– 1198

    2021

  20. [28]

    Connectionist temporal classification: labelling unseg- mented sequence data with recurrent neural networks,

    A. Graves, S. Fern´ andez, F. Gomez, and J. Schmidhuber, “Connectionist temporal classification: labelling unseg- mented sequence data with recurrent neural networks,” in ICML, ser. ICML ’06. NY, USA: Association for Computing Machinery, 2006, p. 369–376. [Online]. Available: ...

    2006

  21. [29]

    Hubert: Self-supervised speech representation learning by masked prediction of hidden units,

    W. Hsu, B. Bolte, Y. Tsai, K. Lakhotia, R. Salakhutdi- nov, and A. Mohamed, “Hubert: Self-supervised speech representation learning by masked prediction of hidden units,” IEEE/ACM Trans. Audio, Speech and Lang. Proc., vol. 29, p. 3451–3460, oct 2021. [Online]. Available: https...

    2021

  22. [30]

    Wavlm: Large-scale self-supervised pre-training for full stack speech processing,

    S. Chen, C. Wang, Z. Chen, Y. Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiao, J. Wu, L. Zhou, S. Ren, Y. Qian, Y. Qian, J. Wu, M. Zeng, X. Yu, and F. Wei, “Wavlm: Large-scale self-supervised pre-training for full stack speech processing,” IEEE Journal of Selected T...

    2022

  23. [31]

    mhubert-147: A compact multilingual hubert model,

    M. Zanon Boito, V. Iyer, N. Lagos, L. Besacier, and I. Calapodescu, “mhubert-147: A compact multilingual hubert model,” in Interspeech 2024, 2024, pp. 3939–3943

    2024

  24. [32]

    A large-scale evaluation of speech foundation models,

    S.-w. Yang, H.-J. Chang, Z. Huang, A. T. Liu, C.-I. Lai, H. Wu, J. Shi, X. Chang, H.-S. Tsai, W.-C. Huang et al., “A large-scale evaluation of speech foundation models,” IEEE/ACM Transactions on Audio, Speech, and Lan- guage Processing, 2024

    2024

  25. [33]

    Cnn-rnn-ctc based end-to-end mispronunciation detection and diagnosis,

    W.-K. Leung, X. Liu, and H. Meng, “Cnn-rnn-ctc based end-to-end mispronunciation detection and diagnosis,” in ICASSP 2019. IEEE, 2019, pp. 8132–8136

    2019

  26. [34]

    Mispronunciation detec- tion and diagnosis in l2 english speech using multidistri- bution deep neural networks,

    K. Li, X. Qian, and H. Meng, “Mispronunciation detec- tion and diagnosis in l2 english speech using multidistri- bution deep neural networks,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 25, no. 1, pp. 193–207, 2016

    2016

Pith tools

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