Advancing Arabic Speech Recognition Through Large-Scale Weakly Supervised Learning

Hasan Abusheikh; Mahmoud Salhab; Marwan Elghitany; Mohammad Abusheikh; Shameed Sait; Syed Sibghat Ullah

REVIEW 4 major objections 5 minor 1 cited by

Advancing Arabic Speech Recognition Through Large-Scale Weakly Supervised Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Arabic speech recognition reaches state-of-the-art with no human transcriptions, using 15,000 hours of weakly supervised audio.

desk verdict A useful Arabic ASR application paper whose SOTA claim currently rests on unverified weak-label quality and test-disjointness; worth refereeing if the authors can produce a decontamination and label audit. read the letter →

arxiv 2504.12254 v2 pith:MYFSYUHW submitted 2025-04-16 cs.AI cs.CL

Mahmoud Salhab , Marwan Elghitany , Shameed Sait , Syed Sibghat Ullah , Mohammad Abusheikh , Hasan Abusheikh This is my paper · ORCID

classification cs.AIcs.CL

keywords AutomaticSpeechRecognitionWeaklySupervisedLearningConformerArabicConnectionistTemporalClassificationWordErrorRateDialectalLow-resourceLanguages

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that high-quality Arabic automatic speech recognition can be built without manually transcribed training data. The authors assemble about 15,000 hours of Arabic speech whose transcriptions are produced by an automated weak-labeling pipeline, then train a Conformer model with connectionist temporal classification loss entirely on those machine-generated labels. On six standard Arabic benchmarks, the model reports the lowest average word error rate, 26.68, and character error rate, 10.05, among the open-source and closed-source systems compared. This matters because manual transcription is the main cost bottleneck for low-resource languages; if machine labels are sufficient, Arabic ASR and similar languages become much cheaper to scale.

What carries the argument

The load-bearing mechanism is the weak-label generation pipeline: speech is segmented by voice activity detection, multiple ASR systems produce candidate transcriptions, the candidate with minimal total Levenshtein distance to the other candidates is chosen, and segments are discarded unless pairwise-average WER/CER and language-model perplexity fall below thresholds; surviving segments are merged into chunks up to 15 seconds. These pseudo-labels train a 121-million-parameter Conformer, a convolution-augmented transformer, with CTC loss and a SentencePiece vocabulary of 1,024 tokens. The pipeline is tuned as an optimization problem on a calibration set to balance annotation yield against transcription error.

What would settle it

Take a few hundred segments from the weakly labeled training set, have native Arabic transcribers produce reference transcripts, and measure the word error rate between the selected hypotheses and these references; if the average is far above the pipeline's own filtering thresholds, the labels are unreliable. As a second check, run a near-duplicate audio search between the internal 30,000-hour corpus and the six benchmark test sets; substantial overlap would invalidate the comparison.

Watch

Extended reading notes

Core claim

The paper's central claim is that weak supervision alone yields state-of-the-art Arabic speech recognition. Starting from an internal corpus of roughly 30,000 unlabeled hours, the labeling pipeline filters and annotates about 15,000 hours by generating multiple hypotheses per segment, keeping only segments where the chosen transcript has low average pairwise word/character error and acceptable language-model perplexity. A Conformer CTC model trained from scratch on these pseudo-labels, decoded greedily without any external language model, achieves an average WER of 26.68 and CER of 10.05 across SADA, Common Voice, MASC clean/noisy, Casablanca, and MGB-2, outperforming every baseline, including a 23.19% average WER reduction over the strongest open-source model.

Load-bearing premise

The state-of-the-art result rests on the assumption that the 15,000 hours of weakly annotated audio are accurate and diverse enough, and sufficiently disjoint from the benchmark test sets, that a model trained on them generalizes to all six test corpora.

Editorial extensions

If this is right

Arabic ASR can be scaled without paid transcriptions: an unlabeled corpus plus automatic labeling directly produces training data.
Other low-resource languages can reuse the same pipeline, provided at least two ASR hypothesis generators exist for them.
A Conformer CTC model with greedy decoding and no language model can outperform larger and closed-source systems, showing that data scale and label quality matter more than decoder complexity.
The two-iteration labeling process suggests that excluding a hallucination-prone hypothesis source improves dataset quality.
The reported 23.19% average WER reduction over the strongest baseline establishes a new quantitative reference point for Arabic ASR evaluation.

Reading between the lines

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

A direct test of the method's core premise would be to human-sample a few hundred weakly labeled segments and measure their actual error rate; the paper does not report such an audit.
The paper never reports a decontamination check between the internal corpus and the test sets, so a near-duplicate analysis is a necessary next step before the benchmark numbers are taken at face value.
If agreement filtering is the true driver of performance, adding more diverse hypothesis generators or dialect-specific models should improve label quality further, which is a testable extension.
The same pipeline could potentially be applied to other low-resource languages that have at least two off-the-shelf ASR systems, reducing the cost of entry for new speech technologies.

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.

Desk Editor's Note

read the letter

The short version: this is a plausible application of known weak-supervision techniques to Arabic ASR, and the reported numbers are better than the listed baselines on most benchmarks. But the headline claim that weak supervision alone achieves state-of-the-art Arabic ASR is not yet supported, because the paper never shows the 15K-hour training corpus is disjoint from the test sets or that the weak labels are accurate enough. I'd send it to peer review with that as the make-or-break condition.

What's actually new: first-to-scale for Arabic with this specific agreement-filtering pipeline, 15K hours, Conformer CTC from scratch, iterated self-training. The evaluation follows the public leaderboard, which is the right thing to do, and the tables are internally consistent. The architecture and training details are fully specified, so replication is possible if data were released.

Soft spots: (1) no decontamination check. Whisper and FastConformer are both trained on massive public data that likely includes Common Voice, MGB-2, etc. Agreement filtering can propagate benchmark text into training labels. (2) No label-error audit on a human-transcribed sample. (3) Eq. 7 optimizes hyperparameters on MASC dev set, so MASC results are partly selection bias. (4) No error bars, so differences like 10.42 vs 10.60 WER on Common Voice are noise-level. (5) Baselines are limited and the closed-source comparison is only three APIs.

These weaknesses are real but not disqualifying if the authors are willing to address them. The core result—that a Conformer trained on filtered pseudo-labels beats Whisper on Arabic—is interesting even without the SOTA framing.

Who should read it: people working on low-resource ASR, especially Arabic; anyone thinking about weak supervision for dialect-rich languages. I would not cite it yet because no data or code, but I would read it. Recommendation: send to peer review with the explicit requirement of a decontamination statement and label audit, or soften the SOTA claim.

Referee Report

4 major / 5 minor

Summary. The paper proposes a weakly supervised pipeline for Arabic ASR. The authors start with roughly 30,000 hours of unlabeled internal audio, segment it, transcribe with a proprietary ASR, FastConformer, and Whisper, filter by average pairwise WER/CER and language-model perplexity, merge segments, and train a 121M-parameter Conformer CTC model from scratch on the resulting roughly 15,000 hours of weak labels. They evaluate on six Arabic benchmarks using the leaderboard protocol of Wang et al. and report the lowest average WER (26.68) and CER (10.05) compared with open and closed-source baselines. The paper does not report a decontamination check between the internal corpus and the test sets, a human label-quality audit, or error bars, and Tables 4-5 contain an internally inconsistent average-reduction calculation.

Significance. If the claims were fully supported, the paper would make a useful contribution: it shows that agreement-filtering of off-the-shelf ASR outputs can produce enough training data for a competitive Arabic ASR model without manual transcription, and it follows a fixed, citable evaluation pipeline with many baselines. The use of a pinned leaderboard commit and the inclusion of both open and closed-source comparisons are strengths. However, the SOTA claim is currently under-supported: the training corpus is not released, its sources are not named, no overlap analysis with the test sets is provided, and the label quality is not measured. The paper would be strengthened by a decontamination statement, a label audit on a human-transcribed sample, and corrected statistics.

major comments (4)

[§3.1, §3.2, §4.3] The central SOTA claim requires that the 15,000-hour weakly labeled corpus used for training is disjoint from the SADA, Common Voice, MASC, MGB-2, and Casablanca test sets, and that the weak labels are sufficiently accurate for a 121M-parameter Conformer trained from scratch. The paper reports neither an audio-level or transcript-level decontamination check nor any human-annotated sample or label-error measurement, and the sources of the internal 30K-hour corpus are never identified. As a result, leakage from the teacher models (Whisper and FastConformer, both of which also appear as baselines) or from the benchmark text itself cannot be ruled out, and the SOTA claim is not yet supported.
[§4.3, Table 5] The statement that the model achieves an average reduction of 23.19% in WER and 24.78% in CER is computed from the aggregate average WER/CER, not from the per-dataset reductions listed in Table 5; averaging the WERR column gives 16.64% and averaging the CERR column gives 15.82%. Moreover, Table 5 reports negative CERR values for Common Voice (-5.24) and MASC clean (-3.01), so the abstract's wording that the model surpasses baselines 'on standard benchmarks' is too strong without qualification.
[§3.2, Eq. (7)] The calibration objective in Eq. (7) tunes the annotation thresholds on the MASC development set, but the text does not state whether any MASC test material was used during development, and the final threshold values (average pairwise WER/CER thresholds, PPL threshold, maximum segment duration, and maximum merged duration) are never reported. Since MASC clean/noisy are two of the six evaluation benchmarks, the paper should clarify the split and report the chosen hyperparameters, otherwise the MASC-family gains may partly reflect tuning on the benchmark rather than weak supervision.
[§4.3] No significance tests, confidence intervals, or multiple-run statistics are reported, yet several of the claimed wins are very small: on Common Voice the WER advantage over Nvidia Conformer is 10.42 vs 10.60, and on CER the model is actually worse on Common Voice and MASC clean. Without uncertainty quantification or at least a paired comparison, the claim that the model 'demonstrably outperforms' all baselines is not established.

minor comments (5)

[§3.2, Algorithm 1] Algorithm 1 appears to place the training step (lines 24-25) inside the per-audio-file loop; if training is meant to happen after all files are processed, the pseudocode should be restructured.
[Abstract, §3.2] There are typographical errors such as 'V oice' in Algorithm 1 and the incomplete sentence in the abstract ('By demonstrating the effectiveness ... paving the way').
[Table 4] Footnote 3 refers to 'SA and JO models' without defining these abbreviations; please expand them.
[§3.1] The paper does not state whether the internal corpus is balanced across the claimed dialects, genders, and age groups; a table with distribution statistics would help.
[§4.1] No code or data release is mentioned; for a claim built on a private 15K-hour corpus, a release plan or at least a detailed data sheet would substantially aid reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported WER/CER come from external human-transcribed benchmarks, and no load-bearing self-citation or definitional reduction is present.

full rationale

The derivation chain is: build a 15k-hour weakly labeled corpus by agreement filtering of FastConformer and Whisper hypotheses (Sections 3.1-3.2), train a Conformer CTC model from scratch on it (Sections 3.3, 4.1), and evaluate WER/CER on six public test sets with human reference transcriptions (Section 4.2). The benchmark numbers are produced by decoding the trained model and comparing against external human-transcribed test sets; they are not computed from, nor defined in terms of, the weak-label generator's outputs. Whisper and FastConformer appear both as label generators and as baselines, but that is a knowledge-distillation/self-training setup, not a definitional reduction: the final WER is an empirical result on held-out data and could in principle be worse than the teachers. Equation (7) tunes pipeline thresholds on the MASC development set, and Table 2 then reports MASC clean/noisy test WER; this is hyperparameter selection on a related but held-out split, not a fit of the reported test statistic. The paper contains no load-bearing self-citations: no reference is authored by the current authors, and no uniqueness theorem is imported. No step reduces to its own input by construction. Therefore no circularity; the unverified-corpus and decontamination concerns are correctness and validity risks, not circularity.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the quality of weakly generated labels, on the absence of overlap between the internal corpus and the benchmark test sets, and on the standard transfer assumption from noisy training labels to clean test distributions. These are domain assumptions, not derived results. The annotation pipeline thresholds are free parameters tuned on MASC development data, but their values are not reported.

free parameters (7)

Average pairwise WER threshold (PWER threshold) = Not disclosed
Segments with average pairwise WER above this threshold are dropped; tuned on MASC dev via Eq. 7 (Section 3.2).
Average pairwise CER threshold (PCER threshold) = Not disclosed
Segments with average pairwise CER above the threshold are dropped; tuned on MASC dev via Eq. 7 (Section 3.2).
Perplexity (PPL) threshold = Not disclosed
Selected hypotheses with PPL above this threshold are discarded; tuned on MASC dev via Eq. 7 (Section 3.2).
Maximum segment duration = 5 seconds
Each audio segment is limited to 5 seconds to minimize data loss if discarded (Section 3.2).
Maximum merged chunk duration = 15 seconds
Consecutive small segments are merged into chunks of at most 15 seconds for training (Section 3.2).
Weight of efficiency minus error in Eq. 7 = 1
The optimization objective uses a unit-weight sum of efficiency minus error; no weighting parameter is reported (Section 3.2).
Number of weak-labeling iterations = 2
The full labeling pipeline was executed over two iterations; more iterations could change the dataset (Section 3.2).

assumptions (3)

domain assumption Test benchmarks are drawn i.i.d. from a clean distribution D_c, while training labels come from a noisy distribution D_n.
This assumption underlies the claim that a model trained on weak labels can generalize to clean test sets (Section 3.1).
domain assumption The internal 30K-hour unlabeled corpus does not overlap with the SADA, Common Voice, MASC, MGB-2, or Casablanca test sets.
No decontamination check is reported, and the SOTA claim requires this assumption to avoid benchmark contamination (Sections 3.1 and 4.3).
domain assumption The agreement-and-perplexity filtering produces weak labels accurate enough to train a strong ASR model.
The paper relies on proxy quality scores instead of human verification; no label-error rate is measured (Section 3.2).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Advancing Arabic Speech Recognition Through Large-Scale Weakly Supervised Learning." pith.science (2026). https://pith.science/paper/MYFSYUHW

@misc{pith2026250412254,
  author       = {Pith},
  title        = {Pith review of: Advancing Arabic Speech Recognition Through Large-Scale Weakly Supervised Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MYFSYUHW}},
  note         = {Machine review of arXiv:2504.12254}
}

read the original abstract

Automatic speech recognition (ASR) is crucial for human-machine interaction in diverse applications like conversational agents, industrial robotics, call center automation, and automated subtitling. However, developing high-performance ASR models remains challenging, particularly for low-resource languages like Arabic, due to the scarcity of large, labeled speech datasets, which are costly and labor-intensive to produce. In this work, we employ weakly supervised learning to train an Arabic ASR model using the Conformer architecture. Our model is trained from scratch on 15,000 hours of weakly annotated speech data covering both Modern Standard Arabic (MSA) and Dialectal Arabic (DA), eliminating the need for costly manual transcriptions. Despite the absence of human-verified labels, our approach achieves state-of-the-art (SOTA) results in Arabic ASR, surpassing both open and closed-source models on standard benchmarks. By demonstrating the effectiveness of weak supervision as a scalable, cost-efficient alternative to traditional supervised approaches, paving the way for improved ASR systems in low resource settings.

Figures

Figures reproduced from arXiv: 2504.12254 by the authors.

Discussion (0). Continue with ORCID 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

NADI 2025: The First Multidialectal Arabic Speech Processing Shared Task
cs.CL 2025-09 conditional novelty 5.0 of 10

The NADI 2025 shared task introduces a standardized speech benchmark for eight Arabic dialects and reports best results of 79.8% dialect ID accuracy, 35.68 WER for ASR, and 55 WER for diacritic restoration.

Reference graph

Works this paper leans on

33 extracted references · 24 canonical work pages · cited by 1 Pith paper

[1]

Arabic speech recognition with deep learning: A review

Wajdan Algihab, Noura Alawwad, Anfal Aldawish, and Sarah AlHumoud. Arabic speech recognition with deep learning: A review. In Gabriele Meiselwitz, editor, Social Computing and Social Media. Design, Human Behavior and Analytics, pages 15–31, Cham, 2019. Springer International Publishing

work page 2019
[2]

Automatic speech recognition using advanced deep learning approaches: A survey

Hamza Kheddar, Mustapha Hemis, and Yassine Himeur. Automatic speech recognition using advanced deep learning approaches: A survey. Information Fusion, page 102422, 2024. 3Given the numerous Arabic Automatic Speech Recognition (ASR) models available on Azure for each dialect, evaluating every model was impractical. Therefore, we selected two representative...

work page 2024
[3]

Industrial applications of automatic speech recognition systems

Jayashri Vajpai and Avnish Bora. Industrial applications of automatic speech recognition systems. International Journal of Engineering Research and Applications, 6(3):88–95, 2016

work page 2016
[4]

A study on speaker-adaptive speech recognition

Xuedong Huang. A study on speaker-adaptive speech recognition. In Speech and Natural Language: Proceedings of a Workshop Held at Pacific Grove, California, February 19-22, 1991, 1991

work page 1991
[5]

On authorship authentication of arabic articles

Ahmed Alwajeeh, Mahmoud Al-Ayyoub, and Ismail Hmeidi. On authorship authentication of arabic articles. In 2014 5th International Conference on Information and Communication Systems (ICICS), pages 1–6, 2014

work page 2014
[6]

Deep learning for arabic nlp: A survey

Mahmoud Al-Ayyoub, Aya Nuseir, Kholoud Alsmearat, Yaser Jararweh, and Brij Gupta. Deep learning for arabic nlp: A survey. Journal of Computational Science, 26:522–531, 2018

work page 2018
[7]

Masc: Massive arabic speech corpus, 2021

Mohammad Al-Fetyani, Muhammad Al-Barham, Gheith Abandah, Adham Alsharkawi, and Maha Dawas. Masc: Massive arabic speech corpus, 2021

work page 2021
[8]

Sada: Saudi audio dataset for arabic

Sadeen Alharbi, Areeb Alowisheq, Zoltán Tüske, Kareem Darwish, Abdullah Alrajeh, Abdulmajeed Alrowithi, Aljawharah Bin Tamran, Asma Ibrahim, Raghad Aloraini, Raneem Alnajim, Ranya Alkahtani, Renad Almuasaad, Sara Alrasheed, Shaykhah Alsubaie, and Yaser Alonaizan. Sada: Saudi audio dataset for arabic. In ICASSP 2024 - 2024 IEEE International Conference on ...

work page 2024

Show all 33 references

[9]

Exploring transformers for large-scale speech recognition

Liang Lu, Changliang Liu, Jinyu Li, and Yifan Gong. Exploring transformers for large-scale speech recognition. arXiv preprint arXiv:2005.09684, 2020

2005 arXiv
[10]

Transformer in action: A comparative study of transformer-based acoustic models for large scale speech recognition applications

Yongqiang Wang, Yangyang Shi, Frank Zhang, Chunyang Wu, Julian Chan, Ching-Feng Yeh, and Alex Xiao. Transformer in action: A comparative study of transformer-based acoustic models for large scale speech recognition applications. In ICASSP 2021 - 2021 IEEE International Confere...

2021
[11]

Learning from flawed data: Weakly supervised automatic speech recognition

Dongji Gao, Hainan Xu, Desh Raj, Leibny Paola Garcia Perera, Daniel Povey, and Sanjeev Khudanpur. Learning from flawed data: Weakly supervised automatic speech recognition. In 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pages 1–8. IEEE, 2023

2023
[12]

Conformer: Convolution-augmented transformer for speech recognition

Anmol Gulati, James Qin, Chung-Cheng Chiu, Niki Parmar, Yu Zhang, Jiahui Yu, Wei Han, Shibo Wang, Zhengdong Zhang, Yonghui Wu, et al. Conformer: Convolution-augmented transformer for speech recognition. arXiv preprint arXiv:2005.08100, 2020

2005 arXiv
[13]

Advances in dialectal arabic speech recognition: A study using twitter to improve egyptian asr

Ahmed Ali, Hamdy Mubarak, and Stephan V ogel. Advances in dialectal arabic speech recognition: A study using twitter to improve egyptian asr. In Proceedings of the 11th International Workshop on Spoken Language Translation: Papers, pages 156–162, 2014

2014
[14]

Hanai, Yu Zhang, James R

Paul Cardinal, Ahmed Ali, Najim Dehak, Yifan Zhang, Takahiro A. Hanai, Yu Zhang, James R. Glass, and Stephan V ogel. Recent advances in asr applied to an arabic transcription system for al-jazeera. InProceedings of Interspeech 2014, pages 2088–2092, 2014

2014
[15]

Morphological decomposition in arabic asr systems

Frank Diehl, Mark JF Gales, Marcus Tomalin, and Philip C Woodland. Morphological decomposition in arabic asr systems. Computer Speech & Language, 26(4):229–243, 2012

2012
[16]

Improving continuous arabic speech recognition over mobile networks dsr and nsr using mfccs features transformed

L Bouchakour and M Debyeche. Improving continuous arabic speech recognition over mobile networks dsr and nsr using mfccs features transformed. International Journal of Circuits, Systems and Signal Processing, 12:1–8, 2018

2018
[17]

Arabic speech recognition by end-to-end, modular systems and human

Amir Hussein, Shinji Watanabe, and Ahmed Ali. Arabic speech recognition by end-to-end, modular systems and human. Computer Speech & Language, 71:101272, 2022

2022
[18]

Morpheme-based feature-rich language models using deep neural networks for lvcsr of egyptian arabic

Amr Mousa, Hong-Kwang Kuo, Lidia Mangu, and Hagen Soltau. Morpheme-based feature-rich language models using deep neural networks for lvcsr of egyptian arabic. pages 8435–8439, 10 2013

2013
[19]

Development of the arabic loria automatic speech recognition system (alasr) and its evaluation for algerian dialect

Mohamed Amine Menacer, Odile Mella, Dominique Fohr, Denis Jouvet, David Langlois, and Kamel Smaïli. Development of the arabic loria automatic speech recognition system (alasr) and its evaluation for algerian dialect. Procedia Computer Science, 117:81–88, 2017. Arabic Computati...

2017
[20]

Darts: Dialectal arabic transcription system

Sameer Khurana, Ahmed Ali, and James Glass. Darts: Dialectal arabic transcription system. arXiv preprint arXiv:1909.12163, 2019

1909 arXiv
[21]

Towards end-to-end speech recognition with recurrent neural networks

Alex Graves and Navdeep Jaitly. Towards end-to-end speech recognition with recurrent neural networks. In International conference on machine learning, pages 1764–1772. PMLR, 2014

2014
[22]

Alsayadi, Abdelaziz A

Hamzah A. Alsayadi, Abdelaziz A. Abdelhamid, Islam Hegazy, and Zaki T. Fayed. Arabic speech recognition using end-to-end deep learning. IET Signal Processing, 15(8):521–534, 2021. 8

2021
[23]

Robust speech recognition via large-scale weak supervision

Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine Mcleavey, and Ilya Sutskever. Robust speech recognition via large-scale weak supervision. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, editors, Proceedin...

2023
[24]

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

Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, and Michael Auli. wav2vec 2.0: A framework for self-supervised learning of speech representations. Advances in neural information processing systems, 33:12449– 12460, 2020

2020
[25]

Deep transfer learning for automatic speech recognition: Towards better generalization

Hamza Kheddar, Yassine Himeur, Somaya Al-Maadeed, Abbes Amira, and Faycal Bensaali. Deep transfer learning for automatic speech recognition: Towards better generalization. Knowledge-Based Systems, 277:110851, 2023

2023
[26]

The people’s speech: A large-scale diverse english speech recognition dataset for commercial usage

Daniel Galvez, Greg Diamos, Juan Ciro, Juan Felipe Cerón, Keith Achorn, Anjali Gopi, David Kanter, Maximilian Lam, Mark Mazumder, and Vijay Janapa Reddi. The people’s speech: A large-scale diverse english speech recognition dataset for commercial usage. arXiv preprint arXiv:21...

2021 arXiv
[27]

Gigaspeech: An evolving, multi-domain asr corpus with 10,000 hours of transcribed audio

Guoguo Chen, Shuzhou Chai, Guanbo Wang, Jiayu Du, Wei-Qiang Zhang, Chao Weng, Dan Su, Daniel Povey, Jan Trmal, Junbo Zhang, et al. Gigaspeech: An evolving, multi-domain asr corpus with 10,000 hours of transcribed audio. arXiv preprint arXiv:2106.06909, 2021

2021 arXiv
[28]

Toward domain-invariant speech recognition via large scale training

Arun Narayanan, Ananya Misra, Khe Chai Sim, Golan Pundak, Anshuman Tripathi, Mohamed Elfeky, Parisa Haghani, Trevor Strohman, and Michiel Bacchiani. Toward domain-invariant speech recognition via large scale training. In 2018 IEEE Spoken Language Technology Workshop (SLT), pag...

2018
[29]

Rethinking evaluation in asr: Are our models robust enough? arXiv preprint arXiv:2010.11745, 2020

Tatiana Likhomanenko, Qiantong Xu, Vineel Pratap, Paden Tomasello, Jacob Kahn, Gilad Avidov, Ronan Collobert, and Gabriel Synnaeve. Rethinking evaluation in asr: Are our models robust enough? arXiv preprint arXiv:2010.11745, 2020

2010 arXiv
[30]

Open universal arabic asr leaderboard

Yingzhi Wang, Anas Alhmoud, and Muhammad Alqurishi. Open universal arabic asr leaderboard. arXiv preprint arXiv:2412.13788, 2024

2024 arXiv
[31]

Ardila, M

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 Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4211–4215, 2020

2020
[32]

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

Ahmed Ali, Peter Bell, James Glass, Yacine Messaoui, Hamdy Mubarak, Steve Renals, and Yifan Zhang. The mgb-2 challenge: Arabic multi-dialect broadcast media recognition. In 2016 IEEE Spoken Language Technology Workshop (SLT), pages 279–284. IEEE, 2016

2016
[33]

Casablanca: Data and models for multidialectal arabic speech recognition

Bashar Talafha, Karima Kadaoui, Samar Mohamed Magdy, Mariem Habiboullah, Chafei Mohamed Chafei, Ahmed Oumar El-Shangiti, Hiba Zayed, Rahaf Alhamouri, Rwaa Assi, Aisha Alraeesi, et al. Casablanca: Data and models for multidialectal arabic speech recognition. arXiv preprint arXi...

arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Arabic speech recognition with deep learning: A review

Wajdan Algihab, Noura Alawwad, Anfal Aldawish, and Sarah AlHumoud. Arabic speech recognition with deep learning: A review. In Gabriele Meiselwitz, editor, Social Computing and Social Media. Design, Human Behavior and Analytics, pages 15–31, Cham, 2019. Springer International Publishing

work page 2019

[2] [2]

Automatic speech recognition using advanced deep learning approaches: A survey

Hamza Kheddar, Mustapha Hemis, and Yassine Himeur. Automatic speech recognition using advanced deep learning approaches: A survey. Information Fusion, page 102422, 2024. 3Given the numerous Arabic Automatic Speech Recognition (ASR) models available on Azure for each dialect, evaluating every model was impractical. Therefore, we selected two representative...

work page 2024

[3] [3]

Industrial applications of automatic speech recognition systems

Jayashri Vajpai and Avnish Bora. Industrial applications of automatic speech recognition systems. International Journal of Engineering Research and Applications, 6(3):88–95, 2016

work page 2016

[4] [4]

A study on speaker-adaptive speech recognition

Xuedong Huang. A study on speaker-adaptive speech recognition. In Speech and Natural Language: Proceedings of a Workshop Held at Pacific Grove, California, February 19-22, 1991, 1991

work page 1991

[5] [5]

On authorship authentication of arabic articles

Ahmed Alwajeeh, Mahmoud Al-Ayyoub, and Ismail Hmeidi. On authorship authentication of arabic articles. In 2014 5th International Conference on Information and Communication Systems (ICICS), pages 1–6, 2014

work page 2014

[6] [6]

Deep learning for arabic nlp: A survey

Mahmoud Al-Ayyoub, Aya Nuseir, Kholoud Alsmearat, Yaser Jararweh, and Brij Gupta. Deep learning for arabic nlp: A survey. Journal of Computational Science, 26:522–531, 2018

work page 2018

[7] [7]

Masc: Massive arabic speech corpus, 2021

Mohammad Al-Fetyani, Muhammad Al-Barham, Gheith Abandah, Adham Alsharkawi, and Maha Dawas. Masc: Massive arabic speech corpus, 2021

work page 2021

[8] [8]

Sada: Saudi audio dataset for arabic

Sadeen Alharbi, Areeb Alowisheq, Zoltán Tüske, Kareem Darwish, Abdullah Alrajeh, Abdulmajeed Alrowithi, Aljawharah Bin Tamran, Asma Ibrahim, Raghad Aloraini, Raneem Alnajim, Ranya Alkahtani, Renad Almuasaad, Sara Alrasheed, Shaykhah Alsubaie, and Yaser Alonaizan. Sada: Saudi audio dataset for arabic. In ICASSP 2024 - 2024 IEEE International Conference on ...

work page 2024