Open Automatic Speech Recognition Models for Classical and Modern Standard Arabic

Boris Ginsburg; Enas Albasiri; Lilit Grigoryan; Nikolay Karpov; Vitaly Lavrukhin

REVIEW 4 major objections 5 minor 47 references

Open Automatic Speech Recognition Models for Classical and Modern Standard Arabic

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read One open data pipeline plus the FastConformer architecture yields an Arabic Modern Standard model with state-of-the-art word error rates and the first open unified model that transcribes Classical Arabic with diacritics.

desk verdict The open Arabic ASR models and pipeline are a real resource, but the SOTA numbers are not trustworthy because the test sets were used to pick the model configuration. read the letter →

arxiv 2507.13977 v1 pith:UOTFZVG6 submitted 2025-07-18 cs.CL eess.AS

Lilit Grigoryan , Nikolay Karpov , Enas Albasiri , Vitaly Lavrukhin , Boris Ginsburg This is my paper · ORCID

classification cs.CLeess.AS

keywords ArabicASRautomaticspeechrecognitionModernStandardClassicaldiacriticspunctuationFastConformeropen-sourcemodels

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 claims that a single universal pipeline for cleaning and normalizing Arabic audio-plus-text, paired with the FastConformer hybrid architecture, is enough to produce open Arabic speech-recognition models that beat prior published systems. Its Modern Standard Arabic model reports word error rates of 8.50% on MASC, 8.20% on Common Voice, and 5.01% on FLEURS, improving on the cited state of the art on all three benchmarks. Its unified MSA-and-Classical-Arabic model is, the authors say, the first open public model to output fully diacritized text for Classical Arabic, reaching a diacritic- and punctuation-aware WER of 6.65% on the overlap-free EveryAyah test split. Both models and the full training recipes are released openly. A reader should care because public Arabic ASR is scarce, Classical Arabic was essentially unserved by end-to-end models, and diacritics carry meaning that plain transcripts drop.

What carries the argument

The load-bearing mechanism is the pairing of the FastConformer hybrid RNN-T/CTC model, with a 1,024-token SentencePiece vocabulary over a 44-symbol diacritized Arabic alphabet, with the paper's data-processing pipeline. The pipeline filters by alphabet and duration, normalizes Eastern Arabic numerals and ligatures, decomposes positional letter forms, removes rare punctuation and Kasheeda, drops samples a pretrained model flags as high-error, and deduplicates texts that appear in both train and test. Its final step, self-filtering the noisy MASC subset to add 255 clean training hours, produces the largest single improvement in the results. The Spanish-pretrained starting checkpoint is a secondary mechanism: it converges faster than from-scratch training, with similar final accuracy.

What would settle it

Retrain the two models while holding out the MASC, MCV, FLEURS, and EveryAyah test splits until the very end, selecting the starting checkpoint and early stopping on a development split, then evaluate once on the held-out sets. If the final MCV WER does not beat the prior 11.7% or the FLEURS WER does not beat 10.30%, the claimed state-of-the-art generalization fails to reproduce; the paper's own from-scratch model (17.53% on MCV, 12.38% on FLEURS) shows how large this selection effect can be.

Watch

Extended reading notes

Core claim

The central discovery is that a FastConformer hybrid RNN-T/CTC model, trained on four open corpora with a 44-symbol alphabet that includes six Hamza forms, six diacritics, and Arabic punctuation, can reach state-of-the-art word error rates on Modern Standard Arabic while also emitting fully diacritized text for Classical Arabic. The MSA model achieves 8.50% WER on MASC, 8.20% on MCV, and 5.01% on FLEURS, beating Whisper-Large-v3, SeamlessM4T, and the previously reported SOTA systems on these sets. Adding the EveryAyah corpus to the training mix yields a unified model whose diacritic- and punctuation-aware WER on the overlap-free EveryAyah test set is 6.65%, a result the paper attributes mainly to the cleaned data rather than to the architecture alone. The authors report that the data pipeline—filtering out-of-alphabet and noisy samples, normalizing numerals and ligatures, removing train/test overlap, and using a pretrained model to extract 255 additional clean hours from the noisy MASC subset—is what moves the combined-set WER from 9.95% to 8.50%.

Load-bearing premise

The load-bearing premise is that the reported word error rates honestly measure generalization to unseen audio, even though the MASC, MCV, and FLEURS test sets were consulted during development to choose the starting checkpoint; if those sets had been reserved for final evaluation only, the new-benchmark claims would be weaker.

Editorial extensions

If this is right

The MSA model outperforms Whisper-Large-v3 and SeamlessM4T on MASC, MCV, and FLEURS, showing that a well-tuned open-data recipe can beat far larger proprietary-trained systems on Arabic.
The unified model is the first open end-to-end ASR system that outputs diacritics for Classical Arabic, with a diacritic-aware WER of 6.65% on the overlap-free EveryAyah split, enabling Qur'anic and classical-text applications such as recitation aid and education.
Punctuation can be added to the output at nearly no cost to word accuracy (8.52% versus 8.50% on MASC), so readability improvements do not trade away recognition quality.
Adding diacritics to the training alphabet has a minor effect on MSA word accuracy, so a single model can serve both fully diacritized Classical Arabic and undiacritized Modern Standard Arabic.
The data-cleaning steps—alphabet filtering, normalization, overlap removal, and noise extraction—transfer to other languages and low-resource settings, since the gains come from data quality rather than architecture.

Reading between the lines

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

The state-of-the-art numbers on MCV and FLEURS are likely optimistic: those test sets were used to pick the Spanish starting checkpoint, so the reported WERs contain selection signal; an honest held-out evaluation would probably land somewhere between the from-scratch numbers (17.53% on MCV, 12.38% on FLEURS) and the reported ones.
The noisy-subset extraction step is effectively a self-training loop, and Table IV shows every cleaning round improves all three benchmarks simultaneously; iterating this loop further is a natural extension that the paper does not explore.
The EveryAyah diacritic results depend on deduplicating train/test overlaps, and the large gap between the 1.55% full-set and 6.65% cleaned-set WERs suggests part of the gain is memorization; the 6.65% figure itself may not transfer to unseen reciters or non-Qur'anic Classical Arabic.
The Spanish-checkpoint warm start hints that the FastConformer family carries phonetically transferable features across languages; a systematic comparison of transfer versus from-scratch training for other low-resource languages would clarify when this recipe is safe to apply.

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.

Desk Editor's Note

read the letter

Hey,

The short version: this paper gives the community something genuinely useful—two open Arabic ASR checkpoints and a real preprocessing pipeline—but the headline SOTA numbers are not trustworthy because the test sets were used to pick the model configuration. That's not a minor quibble; it's the difference between 8.2% and 17.5% on MCV depending on whether you start from Spanish or from scratch.

What's actually new: the first public unified MSA + Classical Arabic model with diacritic and punctuation output, a clean recipe for Arabic data cleaning, and a careful analysis of the available datasets. The CA result on EveryAyah—1.55% diacritic-aware WER after cleaning—looks like a real advance, not just a selection artifact, because it's compared against their own baseline trained without that data. The models are on HuggingFace, which makes the work immediately usable.

Where it falls down: Section IV and Table III compare starting checkpoints on the test sets and pick the winner. Table IV does the same for data-processing choices. That breaks the standard protocol for claiming SOTA. The from-scratch control is worse than the cited SOTA on MCV and FLEURS, so the claimed margins come from choosing a pretrained checkpoint after peeking. Also, the conclusion reports WERs (11.37/9.76/7.73) that don't match Table V's numbers (11.63/10.21/8.18) for the same model. The paper needs to explain that discrepancy. There are no error bars, which for differences of a few tenths of a percent matters.

To be fair, the qualitative conclusion—a Spanish-pretrained checkpoint beats English and multilingual options, and beats prior published results—probably survives, because all three pretrained options beat SOTA. But the exact numbers are optimistically biased and can't be verified from the paper.

Who should read it: anyone working on Arabic ASR or looking for a strong open baseline. It's a useful resource, just not a reliable benchmark paper.

My recommendation: send it to peer review, but require a corrected evaluation protocol (validation-based checkpoint selection, locked test set, consistent numbers) before accepting. The work deserves referee time; the claims just need to be made honest.

Best

Referee Report

4 major / 5 minor

Summary. The paper presents two Arabic automatic speech recognition models built on FastConformer: one for Modern Standard Arabic (MSA) and one unified model covering both MSA and Classical Arabic (CA) with punctuation and diacritics. The models are trained on open datasets (MASC, MCV17.0 Arabic, FLEURS Arabic, and EveryAyah), and the authors describe a data-preprocessing pipeline, report WERs on four evaluation sets, and claim state-of-the-art MSA performance and the first unified public MSA/CA diacritics model. The models and preprocessing code are released to support reproducibility.

Significance. If the reported results are valid, the contributions are practically significant: the paper provides open-source models with a commercial license, a documented preprocessing pipeline, and the first public unified MSA/CA model that outputs diacritics. The MSA model appears to offer large improvements over prior open Arabic ASR systems on MASC and competitive numbers on MCV and FLEURS. However, the central SOTA claim is currently not supported because model selection and data-processing decisions were made using the test sets, and the paper contains conflicting numbers for its headline result. The open-source release and the CA diacritics capability are valuable regardless of the exact SOTA margin.

major comments (4)

[Section IV, Table III] The starting checkpoint is chosen by comparing WERs on the MASC, MCV, and FLEURS test sets, and Table V then reports the Spanish-checkpoint model as 'Our MSA: PC & no Diacr' with the same numbers. This means the headline WERs are the result of selection on the evaluation data, not unbiased estimates of generalization. Because the from-scratch model is substantially worse (17.53% on MCV and 12.38% on FLEURS, versus cited SOTA values of 11.7% and 10.30%), the SOTA claim depends on this selection. Please hold out a validation split for all configuration choices and report final test numbers once, or explicitly present Table III and Table IV as model-selection results and avoid calling the selected numbers 'SOTA'.
[Section VI vs Table V] The conclusion states that the MSA model achieves 11.37% WER on MASC, 9.76% on MCV, and 7.73% on FLEURS, but Table V lists 11.63%, 10.21%, and 8.18% for the same model. These are different numbers, so the paper does not contain a single authoritative statement of its headline result. Please correct the inconsistency and ensure that all sections report the same values.
[Section III.A, Table V] The paper evaluates on 'MASC' sets extracted by the authors from raw data based on video IDs, whereas the cited SOTA of 21.8% [32] is reported on the official MASC test set. Unless the extracted sets exactly match the official MASC test split, the reported improvement on MASC is not directly comparable. Please clarify the relationship between the extracted sets and the official MASC dev/test splits, and if they differ, either provide comparable numbers on the official test set or refrain from a direct SOTA comparison.
[Section V.A, Table IV] The data-preprocessing configurations in Table IV are also compared on the MASC, MCV, and FLEURS test sets, and the final row selects the configuration that yields the best numbers. This is another instance of test-set-based selection, in addition to the checkpoint selection in Table III. The reported improvements from the 'O' preprocessing and the noisy-subset extraction are therefore optimistically biased and should not be treated as final evaluation results.

minor comments (5)

[Table IV] The abbreviation 'extr. nsy' in the last row of Table IV is not defined; please spell it out as 'extracted from noisy subset' or similar.
[Section IV] The statement that 'Best results were obtained with learning rate lr = 5e-3' is given without an ablation or reference; since the learning rate is a free parameter, please provide supporting evidence or a citation.
[Section V.A] The data-filtering loop uses a model trained on clean MASC/MCV/FLEURS data to filter the noisy MASC subset; this is not circular with respect to test labels, but the paper should note that such filtering can propagate systematic transcription errors from the filtering model into the training data.
[Section II.A] Reference [12] is cited as 'Ahmet et al.' in the text, but the reference itself lists the first author as 'A. Ahmed'; please correct the name for consistency.
[Table V] The MediaSpeech column has a footnote about letter normalization applied in post-processing; please clarify whether the same normalization was applied to the prior systems for a fair comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports empirical benchmark results, and no claimed prediction or derivation reduces to its own inputs.

full rationale

The paper's central claims are empirical WER numbers from trained ASR models evaluated on public test sets. No derived quantity is defined in terms of a target result, and no fitted parameter is renamed as a prediction. The data-cleaning procedure trains a model on the clean training subset, uses it to filter noisy MASC training samples, and then retrains; this is a self-training bootstrap on training data, not a circular derivation, and it does not touch test labels. Citations to NVIDIA tools such as NeMo, FastConformer, and the speech-data-processor are infrastructure references, not load-bearing evidence for the Arabic-specific results, and no uniqueness theorem or ansatz is imported from the authors' prior work. The skeptical observation that the Spanish checkpoint was selected using MASC, MCV, and FLEURS test sets is a potential statistical validity issue about reported SOTA margins, not a circularity of the kind where an output is equivalent to an input by construction. Similarly, the inconsistency between Section VI and Table V for the same MSA model is an internal consistency problem, not circular reasoning. The evaluation is against held-out test sets with fixed references, and the reported improvements over baselines are empirical outcomes rather than identities. Therefore the circularity score is 0.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

No new physical or conceptual entities are introduced. The central claim rests on standard ASR assumptions, a set of hand-chosen hyperparameters, and a checkpoint selection procedure that uses test data.

free parameters (7)

Starting checkpoint (Spanish FastConformer) = Spanish checkpoint
Selected based on test-set performance (Table III), violating holdout separation.
Data filtration thresholds = WER > 60, CER > 30
Hand-chosen cutoffs for cleaning the noisy MASC subset (Section V-A).
Learning rate = 5e-3
Tuned for best results (Section IV), chosen after experiments rather than derived.
Warmup steps = 2000
Tuned for best results (Section IV).
CTC weight = 0.3
Hyperparameter for the hybrid RNN-T/CTC decoder (Section IV).
SentencePiece vocabulary size = 1024
Chosen tokenizer size (Section IV).
Number of best checkpoints averaged = 5
Checkpoint averaging policy (Section IV).

assumptions (4)

domain assumption The FastConformer hybrid RNN-T/CTC architecture and its pretrained checkpoints are reliable for Arabic ASR.
The paper builds on NeMo and FastConformer without verifying that this architecture is optimal for Arabic; it relies on prior work (Sections II-B and IV).
domain assumption The extracted MASC dev/test sets correspond to the official MASC splits.
The authors say they used video IDs from the dataset authors, but the exact alignment with the official test set is not independently verified (Section III-A).
domain assumption The overlap removal between training and test sets (MCV17, EveryAyah) eliminates all data leakage.
Duplicate transcripts were removed, but some speaker or audio overlap may remain (Section III-B).
domain assumption MGB2 and MediaSpeech provide a valid out-of-domain blind test.
The models were not trained on these sets, but the distributions may overlap with MASC YouTube content (Section III-A).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Open Automatic Speech Recognition Models for Classical and Modern Standard Arabic." pith.science (2026). https://pith.science/paper/UOTFZVG6

@misc{pith2026250713977,
  author       = {Pith},
  title        = {Pith review of: Open Automatic Speech Recognition Models for Classical and Modern Standard Arabic},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UOTFZVG6}},
  note         = {Machine review of arXiv:2507.13977}
}

read the original abstract

Despite Arabic being one of the most widely spoken languages, the development of Arabic Automatic Speech Recognition (ASR) systems faces significant challenges due to the language's complexity, and only a limited number of public Arabic ASR models exist. While much of the focus has been on Modern Standard Arabic (MSA), there is considerably less attention given to the variations within the language. This paper introduces a universal methodology for Arabic speech and text processing designed to address unique challenges of the language. Using this methodology, we train two novel models based on the FastConformer architecture: one designed specifically for MSA and the other, the first unified public model for both MSA and Classical Arabic (CA). The MSA model sets a new benchmark with state-of-the-art (SOTA) performance on related datasets, while the unified model achieves SOTA accuracy with diacritics for CA while maintaining strong performance for MSA. To promote reproducibility, we open-source the models and their training recipes.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 44 canonical work pages

[32]

Masc: Massive arabic speech corpus,

M. Al-Fetyani, M. Al-Barham, G. Abandah, A. Alsharkawi, and M. Dawas, “Masc: Massive arabic speech corpus,” in 2022 IEEE Spoken Language Technology Workshop (SLT) , 2023, pp. 1006–1013

work page 2022
[1]

Nvidia nemo: A toolkit for conversational ai,

N. Corporation, “Nvidia nemo: A toolkit for conversational ai,” 2020. [Online]. Available: https://developer.nvidia.com/nvidia-nemo

work page 2020
[2]

The development of acoustic models for command and control arabic speech recognition system,

M. Nofal, E. Reheem, H. Henawy, and N. Kader, “The development of acoustic models for command and control arabic speech recognition system,” 01 2004

work page 2004
[3]

A fpga-based hmm for a discrete arabic speech recognition system,

F. Elmisery, A. Khalil, A. Salama, and H. Hammed, “A fpga-based hmm for a discrete arabic speech recognition system,” pp. 322–325, 2003

work page 2003
[4]

New hybrid system (supervised classifier/hmm) for isolated arabic speech recognition,

B. Hocine, R. Djemili, M. Bedda, and C. Snani, “New hybrid system (supervised classifier/hmm) for isolated arabic speech recognition,” 01 2006, pp. 1264 – 1269

work page 2006
[5]

A complete kaldi recipe for building arabic speech recognition systems,

A. Ali, Y . Zhang, P. Cardinal, N. Dahak, S. V ogel, and J. Glass, “A complete kaldi recipe for building arabic speech recognition systems,” 2014 IEEE Workshop on Spoken Language Technology, SLT 2014 - Proceedings, pp. 525–529, 04 2015

work page 2014
[6]

Arabic isolated word speaker dependent recognition system,

A. E. Kourd and K. E. Kourd, “Arabic isolated word speaker dependent recognition system,” Journal of Advances in Mathematics and Computer Science, vol. 14, no. 1, p. 1–15, Jan. 2016

work page 2016
[7]

Recent advances in asr applied to an arabic transcription system for al-jazeera,

P. Cardinal, A. Ali, N. Dehak, Y . Zhang, T. Hanai, J. Glass, and S. V ogel, “Recent advances in asr applied to an arabic transcription system for al-jazeera,” Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH , pp. 2088–2092, 01 2014

work page 2014

Show all 47 references

[8]

Development of the mit asr system for the 2016 arabic multi-genre broadcast challenge,

T. AlHanai, W.-N. Hsu, and J. Glass, “Development of the mit asr system for the 2016 arabic multi-genre broadcast challenge,” in 2016 IEEE Spoken Language Technology Workshop (SLT) , 2016, pp. 299–304

2016
[9]

A comparative study for arabic speech recognition system in noisy environments,

A. Ouisaadane and S. Said, “A comparative study for arabic speech recognition system in noisy environments,” International Journal of Speech Technology, vol. 24, pp. 1–10, 09 2021

2021
[10]

Automatic speech recognition system for tunisian dialect,

A. Masmoudi, F. Bougares, M. Ellouze, Y . Est `eve, and L. Belguith, “Automatic speech recognition system for tunisian dialect,” Language Resources and Evaluation , vol. 52, 03 2018

2018
[11]

Development of a TV broadcasts speech recognition system for qatari Arabic,

M. Elmahdy, M. Hasegawa-Johnson, and E. Mustafawi, “Development of a TV broadcasts speech recognition system for qatari Arabic,” in Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14) , 2014, pp. 3057–3061

2014
[12]

Ahmed, Y

A. Ahmed, Y . Hifny, K. Shaalan, and S. Toral, End-to-End Lexicon Free Arabic Speech Recognition Using Recurrent Neural Networks , 11 2018, pp. 231–248

2018
[13]

Bidirectional deep architecture for arabic speech recognition,

N. Zerari, S. Abdelhamid, H. Bouzgou, and C. Raymond, “Bidirectional deep architecture for arabic speech recognition,” Open Computer Sci- ence, vol. 9, no. 1, pp. 92–102, 2019

2019
[14]

Analyzing phonetic and graphemic representations in end-to-end automatic speech recognition,

Y . Belinkov, A. Ali, and J. Glass, “Analyzing phonetic and graphemic representations in end-to-end automatic speech recognition,” 2020

2020
[15]

Arabic speech recognition using end-to-end deep learning,

H. A. Alsayadi, A. A. Abdelhamid, I. Hegazy, and Z. T. Fayed, “Arabic speech recognition using end-to-end deep learning,” IET Signal Process., vol. 15, pp. 521–534, 2021

2021
[16]

Towards One Model to Rule All: Multilingual Strategy for Dialectal Code-Switching Arabic ASR,

S. A. Chowdhury, A. Hussein, A. Abdelali, and A. Ali, “Towards One Model to Rule All: Multilingual Strategy for Dialectal Code-Switching Arabic ASR,” in Proc. Interspeech 2021 , 2021, pp. 2466–2470

2021
[17]

Aswat: Arabic audio dataset for automatic speech recognition using speech- representation learning,

L. Alkanhal, A. Alessa, E. Almahmoud, and R. Alaqil, “Aswat: Arabic audio dataset for automatic speech recognition using speech- representation learning,” in Proceedings of ArabicNLP 2023 , 2023, pp. 120–127

2023
[18]

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

A. Hussein, S. Watanabe, and A. Ali, “Arabic speech recognition by end-to-end, modular systems and human,” 2021

2021
[19]

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

A. Ali, P. Bell, J. Glass, Y . Messaoui, H. Mubarak, S. Renals, and Y . Zhang, “The mgb-2 challenge: Arabic multi-dialect broadcast media recognition,” 12 2016, pp. 279–284

2016
[20]

A time delay neural net- work architecture for efficient modeling of long temporal contexts,

V . Peddinti, D. Povey, and S. Khudanpur, “A time delay neural net- work architecture for efficient modeling of long temporal contexts,” in Interspeech 2015 , 2015, pp. 3214–3218

2015
[21]

Long short-term memory based recurrent neural network architectures for large vocabulary speech recognition,

H. Sak, A. Senior, and F. Beaufays, “Long short-term memory based recurrent neural network architectures for large vocabulary speech recognition,” 2014

2014
[22]

Connection- ist temporal classification: labelling unsegmented sequence data with recurrent neural networks,

A. Graves, S. Fern ´andez, F. Gomez, and J. Schmidhuber, “Connection- ist temporal classification: labelling unsegmented sequence data with recurrent neural networks,” in Proceedings of the 23rd International Conference on Machine Learning , ser. ICML ’06. New York, NY , USA: ...

2006
[23]

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

A. Hussein, S. Watanabe, and A. Ali, “Arabic speech recognition by end- to-end, modular systems and human,” Computer Speech & Language , vol. 71, p. 101272, 2022

2022
[24]

data2vec: A general framework for self-supervised learning in speech, vision and language,

A. Baevski, W.-N. Hsu, Q. Xu, A. Babu, J. Gu, and M. Auli, “data2vec: A general framework for self-supervised learning in speech, vision and language,” 2022

2022
[25]

Common voice: A massively- multilingual speech corpus,

R. Ardila, M. Branson, K. Davis et al. , “Common voice: A massively- multilingual speech corpus,” in Proceedings of the Twelfth Language Resources and Evaluation Conference , 2020, pp. 4218–4222

2020
[26]

N-Shot Benchmarking of Whisper on Diverse Arabic Speech Recognition,

B. Talafha, A. Waheed, and M. Abdul-Mageed, “N-Shot Benchmarking of Whisper on Diverse Arabic Speech Recognition,” in Proc. INTER- SPEECH 2023 , 2023, pp. 5092–5096

2023
[27]

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

A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak super- vision,” 2022

2022
[28]

Fleurs: Few-shot learning evaluation of universal representations of speech,

A. Conneau, M. Ma, S. Khanuja, Y . Zhang, V . Axelrod, S. Dalmia, J. Riesa, C. Rivera, and A. Bapna, “Fleurs: Few-shot learning evaluation of universal representations of speech,” 2022 IEEE Spoken Language Technology Workshop (SLT), pp. 798–805, 2022

2022
[29]

ArTST: Arabic text and speech transformer,

H. Toyin, A. Djanibekov, A. Kulkarni, and H. Aldarmaki, “ArTST: Arabic text and speech transformer,” in Proceedings of ArabicNLP 2023. Singapore (Hybrid): Association for Computational Linguistics, Dec. 2023, pp. 41–51

2023
[30]

Mediaspeech: Multilanguage asr benchmark and dataset,

R. Kolobov, O. Okhapkina, A. P. Olga Omelchishina, R. Bedyakin, V . Moshkin, D. Menshikov, and N. Mikhaylovskiy, “Mediaspeech: Multilanguage asr benchmark and dataset,” 2021

2021
[31]

Quartznet: Deep automatic speech recognition with 1d time-channel separable convolutions,

S. Kriman, S. Beliaev, B. Ginsburg, J. Huang, O. Kuchaiev, V . Lavrukhin, R. Leary, J. Li, and Y . Zhang, “Quartznet: Deep automatic speech recognition with 1d time-channel separable convolutions,” 2019

2019
[33]

Deep speech: Scaling up end-to-end speech recognition,

A. Hannun, C. Case, J. Casper, B. Catanzaro, G. Diamos, E. Elsen, R. Prenger, S. Satheesh, S. Sengupta, A. Coates, and A. Y . Ng, “Deep speech: Scaling up end-to-end speech recognition,” 2014

2014
[34]

Kaldi speech recognition toolkit,

D. Povey, A. Ghoshal, G. Boulianne, L. Zettlemoyer, V . Panayotov et al. , “Kaldi speech recognition toolkit,” 2011. [Online]. Available: http://kaldi-asr.org/

2011
[35]

Espnet: End-to-end speech processing toolkit,

S. Watanabe, T. Hori, S. Karita, T. Hayashi, J. Nishitoba, Y . Unno, N. E. Y . Soplin, J. Heymann, M. Wiesner, N. Chen, A. Renduchintala, and T. Ochiai, “Espnet: End-to-end speech processing toolkit,” 2018

2018
[36]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proceedings of the 31st International Conference on Neural Information Processing Systems, ser. NIPS’17. Curran Associates Inc., 2017, p. 6000–6010

2017
[37]

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, and R. Pang, “Conformer: Convolution- augmented transformer for speech recognition,” 2020

2020
[38]

Fast conformer with linearly scalable attention for efficient speech recognition,

D. Rekesh, N. R. Koluguri, S. Kriman, S. Majumdar, V . Noroozi, H. Huang, O. Hrinchuk, K. Puvvada, A. Kumar, J. Balam, and B. Gins- burg, “Fast conformer with linearly scalable attention for efficient speech recognition,” 2023

2023
[39]

Tarteel ai’s everyayah dataset,

“Tarteel ai’s everyayah dataset,” https://hf.co/datasets/tarteel-ai/ everyayah, accessed: 2024-09-12

2024
[40]

Speech data processor (sdp) toolkit,

N. Corporation, “Speech data processor (sdp) toolkit,” 2024. [Online]. Available: https://github.com/NVIDIA/NeMo-speech-data-processor

2024
[41]

NeMo (Inverse) Text Normalization: From Development to Production,

Y . Zhang, E. Bakhturina, and B. Ginsburg, “NeMo (Inverse) Text Normalization: From Development to Production,” in Proc. Interspeech 2021, 2021, pp. 4857–4859

2021
[42]

Sequence transduction with recurrent neural networks,

A. Graves, “Sequence transduction with recurrent neural networks,” 2012

2012
[43]

Specaugment: A simple data augmentation method for automatic speech recognition,

D. S. Park, W. Chan, Y . Zhang, C.-C. Chiu, B. Zoph, E. D. Cubuk, and Q. V . Le, “Specaugment: A simple data augmentation method for automatic speech recognition,” in Interspeech 2019. ISCA, Sep. 2019

2019
[44]

Stt en fastconformer hybrid transducer-ctc large p&c,

N. Corporation, “Stt en fastconformer hybrid transducer-ctc large p&c,”
[45]

Stt es fastconformer hybrid transducer-ctc large p&c,

——, “Stt es fastconformer hybrid transducer-ctc large p&c,” 2020. [Online]. Available: https://hf.co/nvidia/stt es fastconformer hybrid large pc

2020
[46]

Seamlessm4t: Massively multilingual & multimodal machine transla- tion,

“Seamlessm4t: Massively multilingual & multimodal machine transla- tion,” 2023

2023
[2020]

Available: https://hf.co/nvidia/stt en fastconformer hybrid large pc

[Online]. Available: https://hf.co/nvidia/stt en fastconformer hybrid large pc

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [32]

Masc: Massive arabic speech corpus,

M. Al-Fetyani, M. Al-Barham, G. Abandah, A. Alsharkawi, and M. Dawas, “Masc: Massive arabic speech corpus,” in 2022 IEEE Spoken Language Technology Workshop (SLT) , 2023, pp. 1006–1013

work page 2022

[2] [1]

Nvidia nemo: A toolkit for conversational ai,

N. Corporation, “Nvidia nemo: A toolkit for conversational ai,” 2020. [Online]. Available: https://developer.nvidia.com/nvidia-nemo

work page 2020

[3] [2]

The development of acoustic models for command and control arabic speech recognition system,

M. Nofal, E. Reheem, H. Henawy, and N. Kader, “The development of acoustic models for command and control arabic speech recognition system,” 01 2004

work page 2004

[4] [3]

A fpga-based hmm for a discrete arabic speech recognition system,

F. Elmisery, A. Khalil, A. Salama, and H. Hammed, “A fpga-based hmm for a discrete arabic speech recognition system,” pp. 322–325, 2003

work page 2003

[5] [4]

New hybrid system (supervised classifier/hmm) for isolated arabic speech recognition,

B. Hocine, R. Djemili, M. Bedda, and C. Snani, “New hybrid system (supervised classifier/hmm) for isolated arabic speech recognition,” 01 2006, pp. 1264 – 1269

work page 2006

[6] [5]

A complete kaldi recipe for building arabic speech recognition systems,

A. Ali, Y . Zhang, P. Cardinal, N. Dahak, S. V ogel, and J. Glass, “A complete kaldi recipe for building arabic speech recognition systems,” 2014 IEEE Workshop on Spoken Language Technology, SLT 2014 - Proceedings, pp. 525–529, 04 2015

work page 2014

[7] [6]

Arabic isolated word speaker dependent recognition system,

A. E. Kourd and K. E. Kourd, “Arabic isolated word speaker dependent recognition system,” Journal of Advances in Mathematics and Computer Science, vol. 14, no. 1, p. 1–15, Jan. 2016

work page 2016

[8] [7]

Recent advances in asr applied to an arabic transcription system for al-jazeera,

P. Cardinal, A. Ali, N. Dehak, Y . Zhang, T. Hanai, J. Glass, and S. V ogel, “Recent advances in asr applied to an arabic transcription system for al-jazeera,” Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH , pp. 2088–2092, 01 2014

work page 2014