---
title: "ADI-20: Arabic Dialect Identification dataset — Elleuch 2025"
description: "Recognising which Arabic dialect a speaker uses is difficult."
authors: ["Haroun Elleuch", "Salima Mdhaffar", "Yannick Esteve", "Fethi Bougares"]
year: 2025
venue: "arXiv:2511.10070"
source: "https://arxiv.org/abs/2511.10070"
doi: "10.21437/interspeech.2025-884"
arxiv: "2511.10070"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/elleuch-2025-adi-20-arabic-dialect-identification/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/elleuch-2025-adi-20-arabic-dialect-identification/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/) / ADI-20: Arabic Dialect Identification dataset and models

# ADI-20: Arabic Dialect Identification dataset and models

**Authors:** [Haroun Elleuch](https://md.research.elamur.ai/free/en/authors/haroun-elleuch/), [Salima Mdhaffar](https://md.research.elamur.ai/free/en/authors/salima-mdhaffar/), [Yannick Esteve](https://md.research.elamur.ai/free/en/authors/yannick-esteve/), [Fethi Bougares](https://md.research.elamur.ai/free/en/authors/fethi-bougares/) · **Year:** 2025 · **Venue:** arXiv:2511.10070 · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://doi.org/10.21437/interspeech.2025-884) · [PDF](https://arxiv.org/pdf/2511.10070) · [DOI 10.21437/interspeech.2025-884](https://doi.org/10.21437/interspeech.2025-884) · [arXiv 2511.10070](https://arxiv.org/abs/2511.10070) · [OpenAlex](https://openalex.org/W4415433505)

**Citations:** 3 citations (openalex-snapshot · 2026-10-14) · [10](#cited-by) in our library, of the 140884 papers whose bibliographies we read

## Summary
Recognising which Arabic dialect a speaker uses is difficult. The varieties differ in sound, vocabulary and grammar, and speakers often switch between dialects, Modern Standard Arabic (MSA) and foreign languages. The earlier ADI-17 corpus had two problems: some dialects had far more data than others, and MSA, Tunisian and Bahraini were absent. The authors build ADI-20, which adds those three classes and tops up the scarcest dialects, giving about 3,556 hours across 19 dialects plus MSA. They then train two kinds of classifiers in SpeechBrain. The first is a pretrained ECAPA-TDNN. The second is a Whisper encoder whose decoder is replaced by attention pooling and a dense output layer. Using duration-capped training subsets and several model sizes, they measure how data volume and parameter count affect accuracy. With only about 30% of the training audio, F1 drops just slightly. At comparable size, ECAPA-TDNN beats Whisper-base. The release of data manifests, recipes and trained models gives the field reproducible baselines, which the authors describe as the first openly available ADI model.

## Key points
- ADI-20 totals 3556.76 hours over 20 classes, and every class has at least 53 training hours.
  - MSA contributes 68 h, drawn from ADI-5 and YouTube news; its dev/test sets are reused from ADI-5.
  - Tunisian contributes 142.7 h, taken from all of TunSwitch including the code-switched part.
  - Bahraini contributes roughly 271 h, all from YouTube.
  - Jordanian and Sudanese were raised to 53 h.
- In ADI-17 the imbalance was severe: Iraqi had 31 times more training audio than Jordanian. Iraqi, Egyptian and Mauritanian together made up more than half the set.
- The training subsets are ADI-17-10h (5.6% of the data), ADI-17-25h (14%), ADI-17-53h (~30%), ADI-17-full, ADI-20-53h and ADI-20-full.
  - Subsets were sampled with one fixed seed and balanced by hours rather than segment counts.
  - Segments were split to at most 30 s, and clips under 3 s were dropped.
- On the ADI-17 test set, weighted F1 was 93.16% for ECAPA-TDNN and 91.34% for the Whisper-base classifier of similar size. The authors link ECAPA-TDNN's edge to its speaker-recognition pretraining.
- The ECAPA-TDNN uses 80-dimensional Mel filterbanks rather than MFCCs and starts from a HuggingFace checkpoint. Training ran on 4 GPUs with batch size 16, and Whisper sizes from tiny to large were explored.

## Interesting ideas
- The authors capped training data per class by total duration, used one fixed seed, and plotted performance against data volume; this showed that a third of the data was nearly enough. — For Mova (learn): Silver treebank subsets could be capped per genre or construction by token count rather than sentence count. This would show where adding more agreed silver data stops paying off.
- They kept only a pretrained encoder, discarded its decoder, and added attention pooling plus a linear head, which turns a generation model into a compact classifier. — For Mova (tree): A small pooled head on the gparse encoder could classify register or source variety before UD dialect conversion. The pooling weights would also expose which tokens drove the decision.
- At equal parameter counts, a model pretrained on a closely related task (speaker recognition) beat a larger-scope multitask model, so the choice of pretraining objective mattered more than generality. — For Mova (learn): When picking starting checkpoints or teacher tiers, compare candidates at a matched parameter budget. Favour the one whose pretraining objective is closest to the target step.
- The standard variety (MSA) was given its own class, separate from regional dialects. This keeps formal speech from being forced into a dialect label. — For Mova (tokens): The tagger could carry an explicit formal/standard register label alongside variety labels such as web or spoken. That way a register mismatch shows up as a visible feature rather than being silently absorbed as noise.
- The authors kept old benchmark test splits unchanged and scored the expanded models on both the old and new test sets, so results stay comparable with earlier work. — For Mova (learn): When the school-test or fable QA suites grow, the previous test sets should be frozen and reported alongside the new ones. This lets regressions be told apart from shifts in difficulty.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

## Cited by

Cited by

10 papers of our library cite it (of the 140884 whose bibliographies we read); 2 of them have a page here, 8 not yet.

- [Ara-Best-RQ: Multi Dialectal Arabic SSL](https://md.research.elamur.ai/free/en/papers/elleuch-2026-ara-best-rq-multi-dialectal/)
- [CTC-DID: CTC-Based Arabic dialect identification for streaming applications](https://md.research.elamur.ai/free/en/papers/farooq-2026-ctc-did-ctc-based-arabic/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{elleuch2025adi,
  title = {{ADI-20: Arabic Dialect Identification dataset and models}},
  author = {Elleuch, Haroun and Mdhaffar, Salima and Esteve, Yannick and Bougares, Fethi},
  year = {2025},
  doi = {10.21437/interspeech.2025-884},
  eprint = {2511.10070},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.21437/interspeech.2025-884}
}
```

- [← previous Adapting Psycholinguistic Research for LLMs: Gender-inclusive Language in a Coreference Context](https://md.research.elamur.ai/free/en/papers/bartl-2025-adapting-psycholinguistic-research-llms-gender/)
- [next → Advancing Arabic Reverse Dictionary Systems: A Transformer-Based Approach with Dataset Construction Guidelines](https://md.research.elamur.ai/free/en/papers/sibaee-2025-advancing-arabic-reverse-dictionary-systems/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/)

## References

29 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. A. G. Pawar and N. V. Patil, “Comparative study of techniques for spoken language dialect identification,” International Journal of Computer Applications, vol. 975, p. 8887, 2024. [find](https://scholar.google.com/scholar?q=A%2E+G%2E+Pawar+and+N%2E+V%2E+Patil%2C+%E2%80%9CComparative+study+of+techniques+for+spoken+language+dialect+identification%2C%E2%80%9D+International+Journal+of+Computer+Applications%2C+vol%2E+975%2C+p%2E+8887%2C+2024%2E)
2. W. Salloum, H. Elfardy, L. Alamir-Salloum, N. Habash, and M. Diab, “Sentence level dialect identification for machine translation system selection,” in Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, 2014. [find](https://scholar.google.com/scholar?q=W%2E+Salloum%2C+H%2E+Elfardy%2C+L%2E+Alamir%2DSalloum%2C+N%2E+Habash%2C+and+M%2E+Diab%2C+%E2%80%9CSentence+level+dialect+identification+for+machine+translation+system+selection%2C%E2%80%9D+in+Proceedings+of+the+52nd+Annual+Meeting+of+the+As)
3. A. Waheed, B. Talafha, P. Sullivan, A. Elmadany, and M. Abdul-Mageed, “VoxArabica: A robust dialect-aware Arabic speech recognition system,” in Proceedings of ArabicNLP 2023. Association for Computational Linguistics, Dec. 2023, pp. 441–449. [find](https://scholar.google.com/scholar?q=A%2E+Waheed%2C+B%2E+Talafha%2C+P%2E+Sullivan%2C+A%2E+Elmadany%2C+and+M%2E+Abdul%2DMageed%2C+%E2%80%9CVoxArabica%3A+A+robust+dialect%2Daware+Arabic+speech+recognition+system%2C%E2%80%9D+in+Proceedings+of+ArabicNLP+2023%2E+Association+for+Computati)
4. A. Kaseb and M. Farouk, “SAIDS: A novel approach for sentiment analysis informed of dialect and sarcasm,” in Proceedings of the Seventh Arabic Natural Language Processing Workshop (WANLP). Association for Computational Linguistics, Dec. 2022, pp. 22–30. [find](https://scholar.google.com/scholar?q=A%2E+Kaseb+and+M%2E+Farouk%2C+%E2%80%9CSAIDS%3A+A+novel+approach+for+sentiment+analysis+informed+of+dialect+and+sarcasm%2C%E2%80%9D+in+Proceedings+of+the+Seventh+Arabic+Natural+Language+Processing+Workshop+%28WANLP%29%2E+Association)
5. K. Doan, A. Waheed, and M. Abdul-Mageed, “Towards zero-shot text-to-speech for Arabic dialects,” in Proceedings of the Second Arabic Natural Language Processing Conference. Association for Computational Linguistics, Aug. 2024, pp. 123–129. [find](https://scholar.google.com/scholar?q=K%2E+Doan%2C+A%2E+Waheed%2C+and+M%2E+Abdul%2DMageed%2C+%E2%80%9CTowards+zero%2Dshot+text%2Dto%2Dspeech+for+Arabic+dialects%2C%E2%80%9D+in+Proceedings+of+the+Second+Arabic+Natural+Language+Processing+Conference%2E+Association+for+Computation)
6. B. Alhafni, S. Al-Towaity, Z. Fawzy, F. Nassar, F. Eryani, H. Bouamor, and N. Habash, “Exploiting dialect identification in automatic dialectal text normalization,” in Proceedings of the Second Arabic Natural Language Processing Conference. Association for Computational Linguistics, Aug. 2024, pp. 42–54. [find](https://scholar.google.com/scholar?q=B%2E+Alhafni%2C+S%2E+Al%2DTowaity%2C+Z%2E+Fawzy%2C+F%2E+Nassar%2C+F%2E+Eryani%2C+H%2E+Bouamor%2C+and+N%2E+Habash%2C+%E2%80%9CExploiting+dialect+identification+in+automatic+dialectal+text+normalization%2C%E2%80%9D+in+Proceedings+of+the+Second+Arabic)
7. C. A. Ferguson, “Diglossia,” in The bilingualism reader. ledge, 2003, pp. 71–86. Rout- [find](https://scholar.google.com/scholar?q=C%2E+A%2E+Ferguson%2C+%E2%80%9CDiglossia%2C%E2%80%9D+in+The+bilingualism+reader%2E+ledge%2C+2003%2C+pp%2E+71%E2%80%9386%2E+Rout%2D)
8. A. Ali, S. Vogel, and S. Renals, “Speech recognition challenge in the wild: Arabic MGB-3,” in 2017 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), Dec. 2017, pp. 316–322. [find](https://scholar.google.com/scholar?q=A%2E+Ali%2C+S%2E+Vogel%2C+and+S%2E+Renals%2C+%E2%80%9CSpeech+recognition+challenge+in+the+wild%3A+Arabic+MGB%2D3%2C%E2%80%9D+in+2017+IEEE+Automatic+Speech+Recognition+and+Understanding+Workshop+%28ASRU%29%2C+Dec%2E+2017%2C+pp%2E+316%E2%80%93322%2E) [our card](https://md.research.elamur.ai/free/en/papers/ali-2017-speech-recognition-challenge-wild-arabic/)
9. A. Kulkarni and H. Aldarmaki, “Yet Another Model for Arabic Dialect Identification,” in Proceedings of ArabicNLP 2023. Association for Computational Linguistics, Dec. 2023, pp. 435–440. [find](https://scholar.google.com/scholar?q=A%2E+Kulkarni+and+H%2E+Aldarmaki%2C+%E2%80%9CYet+Another+Model+for+Arabic+Dialect+Identification%2C%E2%80%9D+in+Proceedings+of+ArabicNLP+2023%2E+Association+for+Computational+Linguistics%2C+Dec%2E+2023%2C+pp%2E+435%E2%80%93440%2E)
10. P. Sullivan, A. Elmadany, and M. Abdul-Mageed, “On the Robustness of Arabic Speech Dialect Identification,” in INTERSPEECH 2023. ISCA, Aug. 2023, pp. 5326–5330. [find](https://scholar.google.com/scholar?q=P%2E+Sullivan%2C+A%2E+Elmadany%2C+and+M%2E+Abdul%2DMageed%2C+%E2%80%9COn+the+Robustness+of+Arabic+Speech+Dialect+Identification%2C%E2%80%9D+in+INTERSPEECH+2023%2E+ISCA%2C+Aug%2E+2023%2C+pp%2E+5326%E2%80%935330%2E) [our card](https://md.research.elamur.ai/free/en/papers/sullivan-2023-robustness-arabic-speech-dialect-identification/)
11. S. Malmasi, M. Zampieri, N. Ljubešić, P. Nakov, A. Ali, and J. Tiedemann, “Discriminating between similar languages and arabic dialect identification: A report on the third dsl shared task,” in Proceedings of the third workshop on NLP for similar languages, varieties and dialects (VarDial3), 2016, pp. 1–14. [find](https://scholar.google.com/scholar?q=S%2E+Malmasi%2C+M%2E+Zampieri%2C+N%2E+Ljubes%CC%8Cic%CC%81%2C+P%2E+Nakov%2C+A%2E+Ali%2C+and+J%2E+Tiedemann%2C+%E2%80%9CDiscriminating+between+similar+languages+and+arabic+dialect+identification%3A+A+report+on+the+third+dsl+shared+task%2C%E2%80%9D+in+Proc) [our card](https://md.research.elamur.ai/free/en/papers/malmasi-2016-discriminating-between-similar-languages/)
12. S. Shon, A. Ali, Y. Samih, H. Mubarak, and J. Glass, “ADI17: A Fine-Grained Arabic Dialect Identification Dataset,” in ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 8244–8248. [find](https://scholar.google.com/scholar?q=S%2E+Shon%2C+A%2E+Ali%2C+Y%2E+Samih%2C+H%2E+Mubarak%2C+and+J%2E+Glass%2C+%E2%80%9CADI17%3A+A+Fine%2DGrained+Arabic+Dialect+Identification+Dataset%2C%E2%80%9D+in+ICASSP+2020+%2D+2020+IEEE+International+Conference+on+Acoustics%2C+Speech+and+Signal+) [our card](https://md.research.elamur.ai/free/en/papers/salameh-2018-fine-grained-arabic-dialect-identification/)
13. A. Ali, S. Shon, Y. Samih, H. Mubarak, A. Abdelali, J. Glass, S. Renals, and K. Choukri, “The MGB-5 Challenge: Recognition and Dialect Identification of Dialectal Arabic Speech,” in 2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), Dec. 2019, pp. 1026–1033. [find](https://scholar.google.com/scholar?q=A%2E+Ali%2C+S%2E+Shon%2C+Y%2E+Samih%2C+H%2E+Mubarak%2C+A%2E+Abdelali%2C+J%2E+Glass%2C+S%2E+Renals%2C+and+K%2E+Choukri%2C+%E2%80%9CThe+MGB%2D5+Challenge%3A+Recognition+and+Dialect+Identification+of+Dialectal+Arabic+Speech%2C%E2%80%9D+in+2019+IEEE+Automati)
14. S. Wray, “Classification of closely related sub-dialects of arabic using support-vector machines,” in Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), 2018. [find](https://scholar.google.com/scholar?q=S%2E+Wray%2C+%E2%80%9CClassification+of+closely+related+sub%2Ddialects+of+arabic+using+support%2Dvector+machines%2C%E2%80%9D+in+Proceedings+of+the+Eleventh+International+Conference+on+Language+Resources+and+Evaluation+%28LREC+20)
15. L. Lulu and A. Elnagar, “Automatic arabic dialect classification using deep learning models,” Procedia computer science, vol. 142, pp. 262–269, 2018. [find](https://scholar.google.com/scholar?q=L%2E+Lulu+and+A%2E+Elnagar%2C+%E2%80%9CAutomatic+arabic+dialect+classification+using+deep+learning+models%2C%E2%80%9D+Procedia+computer+science%2C+vol%2E+142%2C+pp%2E+262%E2%80%93269%2C+2018%2E)
16. W.-N. Hsu, B. Bolte, Y.-H. H. Tsai, K. Lakhotia, R. Salakhutdinov, and A. Mohamed, “Hubert: Self-supervised speech representation learning by masked prediction of hidden units,” IEEE/ACM transactions on audio, speech, and language processing, vol. 29, pp. 3451–3460, 2021. [find](https://scholar.google.com/scholar?q=W%2E%2DN%2E+Hsu%2C+B%2E+Bolte%2C+Y%2E%2DH%2E+H%2E+Tsai%2C+K%2E+Lakhotia%2C+R%2E+Salakhutdinov%2C+and+A%2E+Mohamed%2C+%E2%80%9CHubert%3A+Self%2Dsupervised+speech+representation+learning+by+masked+prediction+of+hidden+units%2C%E2%80%9D+IEEE%2FACM+transactions+) [our card](https://md.research.elamur.ai/free/en/papers/hsu-2021-hubert-self-supervised-speech-representation/)
17. A. Baevski, Y. Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A framework for self-supervised learning of speech representations,” Advances in neural information processing systems, vol. 33, pp. 12 449–12 460, 2020. [find](https://scholar.google.com/scholar?q=A%2E+Baevski%2C+Y%2E+Zhou%2C+A%2E+Mohamed%2C+and+M%2E+Auli%2C+%E2%80%9Cwav2vec+2%2E0%3A+A+framework+for+self%2Dsupervised+learning+of+speech+representations%2C%E2%80%9D+Advances+in+neural+information+processing+systems%2C+vol%2E+33%2C+pp%2E+12+449%E2%80%93) [our card](https://md.research.elamur.ai/free/en/papers/baevski-2020-wav2vec-2-0-framework-self/)
18. A. Djanibekov, H. O. Toyin, R. Alshalan, A. Alitr, and H. Aldarmaki, “Dialectal coverage and generalization in arabic speech recognition,” arXiv preprint arXiv:2411.05872, 2024. [↗](https://arxiv.org/abs/2411.05872)
19. S. Mdhaffar, F. Bougares, R. de Mori, S. Zaiem, M. Ravanelli, and Y. Estève, “TARIC-SLU: A Tunisian benchmark dataset for spoken language understanding,” in LREC-COLING 2024. Torino, Italia: ELRA and ICCL, May 2024, pp. 15 606–15 616. [Online]. Available: https://aclanthology.org/2024.lrec-main.1357/ [↗](https://aclanthology.org/2024.lrec-main.1357/)
20. S. Mdhaffar, H. Elleuch, F. Bougares, and Y. Estève, “Performance analysis of speech encoders for low-resource slu and asr in tunisian dialect,” in Proceedings of The Second Arabic Natural Language Processing Conference, 2024, pp. 130–139. [find](https://scholar.google.com/scholar?q=S%2E+Mdhaffar%2C+H%2E+Elleuch%2C+F%2E+Bougares%2C+and+Y%2E+Este%CC%80ve%2C+%E2%80%9CPerformance+analysis+of+speech+encoders+for+low%2Dresource+slu+and+asr+in+tunisian+dialect%2C%E2%80%9D+in+Proceedings+of+The+Second+Arabic+Natural+Language+P)
21. A. A. B. Abdallah, A. Kabboudi, A. Kanoun, and S. Zaiem, “Leveraging Data Collection and Unsupervised Learning for Code-Switched Tunisian Arabic Automatic Speech Recognition,” in ICASSP, Apr. 2024, pp. 12 607–12 611. [find](https://scholar.google.com/scholar?q=A%2E+A%2E+B%2E+Abdallah%2C+A%2E+Kabboudi%2C+A%2E+Kanoun%2C+and+S%2E+Zaiem%2C+%E2%80%9CLeveraging+Data+Collection+and+Unsupervised+Learning+for+Code%2DSwitched+Tunisian+Arabic+Automatic+Speech+Recognition%2C%E2%80%9D+in+ICASSP%2C+Apr%2E+2024%2C+pp) [our card](https://md.research.elamur.ai/free/en/papers/abdallah-2023-leveraging-data-collection-unsupervised-learning/)
22. M. Ravanelli, T. Parcollet, A. Moumen, S. de Langen, C. Subakan, P. Plantinga, Y. Wang, P. Mousavi, L. D. Libera, A. Ploujnikov, F. Paissan, D. Borra, S. Zaiem, Z. Zhao, S. Zhang, G. Karakasidis, S.-L. Yeh, P. Champion, A. Rouhe, R. Braun, F. Mai, J. Zuluaga-Gomez, S. M. Mousavi, A. Nautsch, H. Nguyen, X. Liu, S. Sagar, J. Duret, S. Mdhaffar, G. Laperrière, M. Rouvier, R. D. Mori, and Y. Estève, “Open-source conversational AI with SpeechBrain 1.0,” Journal of Machine Learning Research, 2024. [find](https://scholar.google.com/scholar?q=M%2E+Ravanelli%2C+T%2E+Parcollet%2C+A%2E+Moumen%2C+S%2E+de+Langen%2C+C%2E+Subakan%2C+P%2E+Plantinga%2C+Y%2E+Wang%2C+P%2E+Mousavi%2C+L%2E+D%2E+Libera%2C+A%2E+Ploujnikov%2C+F%2E+Paissan%2C+D%2E+Borra%2C+S%2E+Zaiem%2C+Z%2E+Zhao%2C+S%2E+Zhang%2C+G%2E+Karakasidis%2C+S%2E%2DL)
23. B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification,” in Interspeech 2020. ISCA, Oct. 2020, pp. 3830–3834. [find](https://scholar.google.com/scholar?q=B%2E+Desplanques%2C+J%2E+Thienpondt%2C+and+K%2E+Demuynck%2C+%E2%80%9CECAPA%2DTDNN%3A+Emphasized+Channel+Attention%2C+Propagation+and+Aggregation+in+TDNN+Based+Speaker+Verification%2C%E2%80%9D+in+Interspeech+2020%2E+ISCA%2C+Oct%2E+2020%2C+pp%2E+38)
24. A. Radford, J. W. Kim, T. Xu, G. Brockman, C. Mcleavey, and I. Sutskever, “Robust Speech Recognition via Large-Scale Weak Supervision,” in Proceedings of the 40th International Conference on Machine Learning. PMLR, Jul. 2023, pp. 28 492–28 518. [find](https://scholar.google.com/scholar?q=A%2E+Radford%2C+J%2E+W%2E+Kim%2C+T%2E+Xu%2C+G%2E+Brockman%2C+C%2E+Mcleavey%2C+and+I%2E+Sutskever%2C+%E2%80%9CRobust+Speech+Recognition+via+Large%2DScale+Weak+Supervision%2C%E2%80%9D+in+Proceedings+of+the+40th+International+Conference+on+Machine+L) [our card](https://md.research.elamur.ai/free/en/papers/radford-2022-robust-speech-recognition-large-scale/)
25. B. Talafha, A. Waheed, and M. Abdul-Mageed, “N-Shot Benchmarking of Whisper on Diverse Arabic Speech Recognition,” in INTERSPEECH 2023. ISCA, Aug. 2023, pp. 5092–5096. [find](https://scholar.google.com/scholar?q=B%2E+Talafha%2C+A%2E+Waheed%2C+and+M%2E+Abdul%2DMageed%2C+%E2%80%9CN%2DShot+Benchmarking+of+Whisper+on+Diverse+Arabic+Speech+Recognition%2C%E2%80%9D+in+INTERSPEECH+2023%2E+ISCA%2C+Aug%2E+2023%2C+pp%2E+5092%E2%80%935096%2E) [our card](https://md.research.elamur.ai/free/en/papers/talafha-2023-n-shot-benchmarking-whisper-diverse/)
26. A. Babu, C. Wang, A. Tjandra, K. Lakhotia, Q. Xu, N. Goyal, K. Singh, P. Von Platen, Y. Saraf, J. Pino, A. Baevski, A. Conneau, and M. Auli, “XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale,” in Interspeech 2022. ISCA, Sep. 2022, pp. 2278–2282. [find](https://scholar.google.com/scholar?q=A%2E+Babu%2C+C%2E+Wang%2C+A%2E+Tjandra%2C+K%2E+Lakhotia%2C+Q%2E+Xu%2C+N%2E+Goyal%2C+K%2E+Singh%2C+P%2E+Von+Platen%2C+Y%2E+Saraf%2C+J%2E+Pino%2C+A%2E+Baevski%2C+A%2E+Conneau%2C+and+M%2E+Auli%2C+%E2%80%9CXLS%2DR%3A+Self%2Dsupervised+Cross%2Dlingual+Speech+Representation) [our card](https://md.research.elamur.ai/free/en/papers/babu-2021-xls-r-self-supervised-cross/)
27. B. Talafha, K. Kadaoui, S. M. Magdy, M. Abdul-Mageed et al., “Casablanca: Data and Models for Multidialectal Arabic Speech Recognition,” in Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, Nov. 2024, pp. 21 745–21 758. [find](https://scholar.google.com/scholar?q=B%2E+Talafha%2C+K%2E+Kadaoui%2C+S%2E+M%2E+Magdy%2C+M%2E+Abdul%2DMageed+et+al%2E%2C+%E2%80%9CCasablanca%3A+Data+and+Models+for+Multidialectal+Arabic+Speech+Recognition%2C%E2%80%9D+in+Proceedings+of+the+2024+Conference+on+Empirical+Methods+in+N)
28. O. Eberhard and T. Zesch, “Effects of layer freezing on transferring a speech recognition system to under-resourced languages,” in KONVENS 2021, K. Evang, L. Kallmeyer, R. Osswald, J. Waszczuk, and T. Zesch, Eds. KONVENS 2021 Organizers, 6–9 Sep. 2021, pp. 208–212. [find](https://scholar.google.com/scholar?q=O%2E+Eberhard+and+T%2E+Zesch%2C+%E2%80%9CEffects+of+layer+freezing+on+transferring+a+speech+recognition+system+to+under%2Dresourced+languages%2C%E2%80%9D+in+KONVENS+2021%2C+K%2E+Evang%2C+L%2E+Kallmeyer%2C+R%2E+Osswald%2C+J%2E+Waszczuk%2C+and+T%2E)
29. M. Zanon Boito, J. Ortega, H. Riguidel, A. Laurent, L. Barrault, F. Bougares, F. Chaabani, H. Nguyen, F. Barbier, S. Gahbiche, and Y. Estève, “ON-TRAC consortium systems for the IWSLT 2022 dialect and low-resource speech translation tasks,” in IWSLT 2022, E. Salesky, M. Federico, and M. Costa-jussà, Eds. Association for Computational Linguistics, May 2022, pp. 308–318. [find](https://scholar.google.com/scholar?q=M%2E+Zanon+Boito%2C+J%2E+Ortega%2C+H%2E+Riguidel%2C+A%2E+Laurent%2C+L%2E+Barrault%2C+F%2E+Bougares%2C+F%2E+Chaabani%2C+H%2E+Nguyen%2C+F%2E+Barbier%2C+S%2E+Gahbiche%2C+and+Y%2E+Este%CC%80ve%2C+%E2%80%9CON%2DTRAC+consortium+systems+for+the+IWSLT+2022+dialect+an)