---
title: "Arabic Code-Switching Speech Recognition using — Ali 2021"
description: "The paper tackles Arabic–English code-switching in speech recognition when no mixed-language training material exists."
authors: ["Ahmed Ali", "Shammur Chowdhury", "Amir Hussein", "Yasser Hifny"]
year: 2021
venue: "arXiv:2107.01573"
source: "https://arxiv.org/abs/2107.01573"
doi: "10.48550/arxiv.2107.01573"
arxiv: "2107.01573"
licence: "CC BY-NC-SA 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by-nc-sa/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/ali-2021-arabic-code-switching-speech-recognition/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/ali-2021-arabic-code-switching-speech-recognition/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2021](https://md.research.elamur.ai/free/en/papers/2021/) / Arabic Code-Switching Speech Recognition using Monolingual Data

# Arabic Code-Switching Speech Recognition using Monolingual Data

**Authors:** [Ahmed Ali](https://md.research.elamur.ai/free/en/authors/ahmed-ali/), [Shammur Chowdhury](https://md.research.elamur.ai/free/en/authors/shammur-chowdhury/), [Amir Hussein](https://md.research.elamur.ai/free/en/authors/amir-hussein/), [Yasser Hifny](https://md.research.elamur.ai/free/en/authors/yasser-hifny/) · **Year:** 2021 · **Venue:** arXiv:2107.01573 · **Licence:** [CC BY-NC-SA 4.0 (arXiv)](http://creativecommons.org/licenses/by-nc-sa/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2107.01573) · [PDF](https://arxiv.org/pdf/2107.01573) · [DOI 10.48550/arxiv.2107.01573](https://doi.org/10.48550/arxiv.2107.01573) · [arXiv 2107.01573](https://arxiv.org/abs/2107.01573) · [OpenAlex](https://openalex.org/W3181647671)

**Citations:** 15 citations (openalex-snapshot · 2026-10-14) · [12](#cited-by) in our library, of the 140884 papers whose bibliographies we read

## Summary
The paper tackles Arabic–English code-switching in speech recognition when no mixed-language training material exists. Only monolingual corpora are used: MGB-2 for Arabic and TED-LIUM3 for English. The authors build hybrid Kaldi systems with a TDNN acoustic model trained with LF-MMI over a pooled grapheme inventory in which each grapheme carries its language tag. They test two ways of producing a single decoding graph:

- **Global G (baseline):** one 4-gram language model trained on the concatenated monolingual texts.
- **Multi-graph (new):** a separate HCLG graph per language, joined by a union. A Kleene closure then links final states back to the start, so the search can move into the other language at any word boundary.

Both hybrids are compared with a transformer encoder–decoder trained on the same data. The WFST decoders handle switches that happen between sentences better. The transformer is stronger when the switch happens inside a sentence. The authors also publish synthetic dev/test sets and a naturally occurring (ecological) code-switching test set. This addresses the lack of reproducible public benchmarks for Arabic–English code-switching.

## Key points
- Training data: MGB-2 Arabic gives 1,200h (370K segments) and TED-LIUM3 English gives 450h (268K segments). Dev/test sets are 10h/10h for Arabic and 1.6h/2.6h for English.
- Hybrid acoustic model: TDNN trained with LF-MMI on GMM-HMM alignments. Each frame's input is five spliced 40-dim MFCC vectors plus a 100-dim i-vector, giving 300 dimensions.
- Multi-graph build: each language's HCLG is composed with a sigma matcher FST to remove epsilon-only paths, which otherwise cause Kaldi decoding errors. The graphs are then combined by union and closure, so no code-switched LM text is needed.
- End-to-end baseline: transformer with BPE units (size 5000), trained on a joint CTC + cross-entropy objective.
- Main finding: WFST approaches suit inter-sentential switching, while the transformer suits intra-sentential switching. The new test sets are released for benchmarking.

## Interesting ideas
- Instead of training one model on mixed data, compile one automaton per "language" and let the search hop between them through union plus closure, so switching emerges without any mixed examples. — For Mova (tokens): the tokenizer could run several sub-lexers (plain English, numbers/units, URLs/code, foreign names) as a union with closure, and label each span with the sub-lexer that produced it, which keeps the segmentation explainable.
- When pooling units from different sources into one inventory, keep the source identity attached to each unit rather than merging look-alikes. — For Mova (world): when compiling WordNet, FrameNet and Wikidata into one ontology, tag every node and edge with its origin so conflicting merges can be traced and undone.
- Splitting evaluation by phenomenon type (switches between sentences vs. within a sentence) showed that different architectures win on different slices, which a single overall score would hide. — For Mova (learn): report the perceptron parser and gparse separately on slices such as cross-sentence vs. within-sentence coreference or long vs. short dependencies, to decide which witness to trust where.
- Pair an artificially constructed test set with a natural "ecological" one, so gains on clean synthetic data can be checked against real usage. — For Mova (answer): alongside generated maths word problems and templated QA, keep a small set of naturally written school tests as a reality check that synthetic improvements transfer.

*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

12 papers of our library cite it (of the 140884 whose bibliographies we read); 3 of them have a page here, 9 not yet.

- [Fanar: An Arabic-Centric Multimodal Generative AI Platform](https://md.research.elamur.ai/free/en/papers/abbas-2025-fanar-arabic-centric-multimodal-generative/)
- [Hierarchical Aggregation of Dialectal Data for Arabic Dialect Identification](https://md.research.elamur.ai/free/en/papers/baimukan-2022-hierarchical-aggregation-dialectal-data-arabic/)
- [Towards One Model to Rule All: Multilingual Strategy for Dialectal Code-Switching Arabic ASR](https://md.research.elamur.ai/free/en/papers/chowdhury-2021-towards-one-model-rule-all/)

## Cite

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

```bibtex
@misc{ali2021arabic,
  title = {{Arabic Code-Switching Speech Recognition using Monolingual Data}},
  author = {Ali, Ahmed and Chowdhury, Shammur and Hussein, Amir and Hifny, Yasser},
  year = {2021},
  doi = {10.48550/arxiv.2107.01573},
  eprint = {2107.01573},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2107.01573}
}
```

- [← previous Arabic aspect sentiment polarity classification using BERT](https://md.research.elamur.ai/free/en/papers/m-2021-abdelgwad-arabic-aspect-sentiment-polarity-classification/)
- [next → Arabic Offensive Language on Twitter: Analysis and Experiments](https://md.research.elamur.ai/free/en/papers/mubarak-2021-arabic-offensive-language-twitter-analysis/)
- [↑ back to papers of 2021](https://md.research.elamur.ai/free/en/papers/2021/)

## References

36 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. H. Lin, L. Deng, D. Yu, Y. Gong, A. Acero, and C. Lee, “A study on multilingual acoustic modeling for large vocabulary ASR,” in ICASSP, 2009. [find](https://scholar.google.com/scholar?q=H%2E+Lin%2C+L%2E+Deng%2C+D%2E+Yu%2C+Y%2E+Gong%2C+A%2E+Acero%2C+and+C%2E+Lee%2C+%E2%80%9CA+study+on+multilingual+acoustic+modeling+for+large+vocabulary+ASR%2C%E2%80%9D+in+ICASSP%2C+2009%2E)
2. J. Huang, J. Li, D. Yu, L. Deng, and Y. Gong, “Cross-language knowledge transfer using multilingual deep neural network with shared hidden layers,” in 2013 IEEE International Conference on Acoustics, Speech and Signal Processing, 2013. [find](https://scholar.google.com/scholar?q=J%2E+Huang%2C+J%2E+Li%2C+D%2E+Yu%2C+L%2E+Deng%2C+and+Y%2E+Gong%2C+%E2%80%9CCross%2Dlanguage+knowledge+transfer+using+multilingual+deep+neural+network+with+shared+hidden+layers%2C%E2%80%9D+in+2013+IEEE+International+Conference+on+Acoustics%2C+)
3. G. Heigold, V. Vanhoucke, A. Senior, P. Nguyen, M. Ranzato, M. Devin, and J. Dean, “Multilingual acoustic models using distributed deep neural networks,” in ICASSP, 2013. [find](https://scholar.google.com/scholar?q=G%2E+Heigold%2C+V%2E+Vanhoucke%2C+A%2E+Senior%2C+P%2E+Nguyen%2C+M%2E+Ranzato%2C+M%2E+Devin%2C+and+J%2E+Dean%2C+%E2%80%9CMultilingual+acoustic+models+using+distributed+deep+neural+networks%2C%E2%80%9D+in+ICASSP%2C+2013%2E)
4. L. Burget, P. Schwarz, M. Agarwal, P. Akyazi, K. Feng, A. Ghoshal, O. Glembek, N. Goel, M. Karafiát, D. Povey, A. Rastrow, R. C. Rose, and S. Thomas, “Multilingual acoustic modeling for speech recognition based on subspace gaussian mixture models,” in ICASSP, 2010. [find](https://scholar.google.com/scholar?q=L%2E+Burget%2C+P%2E+Schwarz%2C+M%2E+Agarwal%2C+P%2E+Akyazi%2C+K%2E+Feng%2C+A%2E+Ghoshal%2C+O%2E+Glembek%2C+N%2E+Goel%2C+M%2E+Karafia%CC%81t%2C+D%2E+Povey%2C+A%2E+Rastrow%2C+R%2E+C%2E+Rose%2C+and+S%2E+Thomas%2C+%E2%80%9CMultilingual+acoustic+modeling+for+speech+recogn)
5. S. Tong, P. N. Garner, and H. Bourlard, “An investigation of multilingual ASR using end-to-end LF-MMI,” in ICASSP, 2019. [find](https://scholar.google.com/scholar?q=S%2E+Tong%2C+P%2E+N%2E+Garner%2C+and+H%2E+Bourlard%2C+%E2%80%9CAn+investigation+of+multilingual+ASR+using+end%2Dto%2Dend+LF%2DMMI%2C%E2%80%9D+in+ICASSP%2C+2019%2E)
6. S. Toshniwal, T. N. Sainath, R. J. Weiss, B. Li, P. Moreno, E. Weinstein, and K. Rao, “Multilingual speech recognition with a single end-to-end model,” in ICASSP, 2018. [find](https://scholar.google.com/scholar?q=S%2E+Toshniwal%2C+T%2E+N%2E+Sainath%2C+R%2E+J%2E+Weiss%2C+B%2E+Li%2C+P%2E+Moreno%2C+E%2E+Weinstein%2C+and+K%2E+Rao%2C+%E2%80%9CMultilingual+speech+recognition+with+a+single+end%2Dto%2Dend+model%2C%E2%80%9D+in+ICASSP%2C+2018%2E)
7. A. Datta, B. Ramabhadran, J. Emond, A. Kannan, and B. Roark, “Language-agnostic multilingual modeling,” arXiv preprint arXiv:2004.09571, 2020. [↗](https://arxiv.org/abs/2004.09571)
8. S. Sitaram, K. R. Chandu, S. K. Rallabandi, and A. W. Black, “A survey of code-switched speech and language processing,” arXiv preprint arXiv:1904.00784, 2019. [↗](https://arxiv.org/abs/1904.00784)
9. Y. Li and P. Fung, “Improved mixed language speech recognition using asymmetric acoustic model and language model with code-switch inversion constraints,” in ICASSP, 2013. [find](https://scholar.google.com/scholar?q=Y%2E+Li+and+P%2E+Fung%2C+%E2%80%9CImproved+mixed+language+speech+recognition+using+asymmetric+acoustic+model+and+language+model+with+code%2Dswitch+inversion+constraints%2C%E2%80%9D+in+ICASSP%2C+2013%2E)
10. ——, “Language modeling for mixed language speech recognition using weighted phrase extraction.” in Interspeech, 2013. [find](https://scholar.google.com/scholar?q=%E2%80%94%E2%80%94%2C+%E2%80%9CLanguage+modeling+for+mixed+language+speech+recognition+using+weighted+phrase+extraction%2E%E2%80%9D+in+Interspeech%2C+2013%2E)
11. N. T. Vu, D.-C. Lyu, J. Weiner, D. Telaar, T. Schlippe, F. Blaicher, E.-S. Chng, T. Schultz, and H. Li, “A first speech recognition system for Mandarin-English code-switch conversational speech,” in ICASSP, 2012. [find](https://scholar.google.com/scholar?q=N%2E+T%2E+Vu%2C+D%2E%2DC%2E+Lyu%2C+J%2E+Weiner%2C+D%2E+Telaar%2C+T%2E+Schlippe%2C+F%2E+Blaicher%2C+E%2E%2DS%2E+Chng%2C+T%2E+Schultz%2C+and+H%2E+Li%2C+%E2%80%9CA+first+speech+recognition+system+for+Mandarin%2DEnglish+code%2Dswitch+conversational+speech%2C%E2%80%9D+in+I)
12. E. Yilmaz, M. Andringa, S. Kingma, J. Dijkstra, F. Kuip, H. Velde, F. Kampstra, J. Algra, H. Heuvel, and D. A. van Leeuwen, “A longitudinal bilingual frisian-dutch radio broadcast database designed for code-switching research,” 2016. [find](https://scholar.google.com/scholar?q=E%2E+Yilmaz%2C+M%2E+Andringa%2C+S%2E+Kingma%2C+J%2E+Dijkstra%2C+F%2E+Kuip%2C+H%2E+Velde%2C+F%2E+Kampstra%2C+J%2E+Algra%2C+H%2E+Heuvel%2C+and+D%2E+A%2E+van+Leeuwen%2C+%E2%80%9CA+longitudinal+bilingual+frisian%2Ddutch+radio+broadcast+database+designed+fo)
13. A. Pandey, B. M. L. Srivastava, R. Kumar, B. T. Nellore, K. S. Teja, and S. V. Gangashetty, “Phonetically balanced code-mixed speech corpus for Hindi-English automatic speech recognition,” in LREC 2018, 2018. [find](https://scholar.google.com/scholar?q=A%2E+Pandey%2C+B%2E+M%2E+L%2E+Srivastava%2C+R%2E+Kumar%2C+B%2E+T%2E+Nellore%2C+K%2E+S%2E+Teja%2C+and+S%2E+V%2E+Gangashetty%2C+%E2%80%9CPhonetically+balanced+code%2Dmixed+speech+corpus+for+Hindi%2DEnglish+automatic+speech+recognition%2C%E2%80%9D+in+LREC+201)
14. K. Taneja, S. Guha, P. Jyothi, and B. Abraham, “Exploiting monolingual speech corpora for code-mixed speech recognition,” Interspeech, 2019. [find](https://scholar.google.com/scholar?q=K%2E+Taneja%2C+S%2E+Guha%2C+P%2E+Jyothi%2C+and+B%2E+Abraham%2C+%E2%80%9CExploiting+monolingual+speech+corpora+for+code%2Dmixed+speech+recognition%2C%E2%80%9D+Interspeech%2C+2019%2E)
15. G. Sreeram and R. Sinha, “Exploration of end-to-end framework for code-switching speech recognition task: Challenges and enhancements,” IEEE Access, vol. 8, pp. 68 146–68 157, 2020. [find](https://scholar.google.com/scholar?q=G%2E+Sreeram+and+R%2E+Sinha%2C+%E2%80%9CExploration+of+end%2Dto%2Dend+framework+for+code%2Dswitching+speech+recognition+task%3A+Challenges+and+enhancements%2C%E2%80%9D+IEEE+Access%2C+vol%2E+8%2C+pp%2E+68+146%E2%80%9368+157%2C+2020%2E)
16. D. Amazouz, M. Adda-Decker, and L. Lamel, “Addressing code-switching in French/Algerian Arabic speech,” in Interspeech, 2017. [find](https://scholar.google.com/scholar?q=D%2E+Amazouz%2C+M%2E+Adda%2DDecker%2C+and+L%2E+Lamel%2C+%E2%80%9CAddressing+code%2Dswitching+in+French%2FAlgerian+Arabic+speech%2C%E2%80%9D+in+Interspeech%2C+2017%2E)
17. A. Ali, S. Chowdhury, M. Afify, W. El-Hajj, H. Hajj, M. Abbas, A. Hussein, N. Ghneim, M. Abushariah, and A. Alqudah, “Connecting arabs: bridging the gap in dialectal speech recognition,” Communications of the ACM, vol. 64, no. 4, pp. 124–129, 2021. [find](https://scholar.google.com/scholar?q=A%2E+Ali%2C+S%2E+Chowdhury%2C+M%2E+Afify%2C+W%2E+El%2DHajj%2C+H%2E+Hajj%2C+M%2E+Abbas%2C+A%2E+Hussein%2C+N%2E+Ghneim%2C+M%2E+Abushariah%2C+and+A%2E+Alqudah%2C+%E2%80%9CConnecting+arabs%3A+bridging+the+gap+in+dialectal+speech+recognition%2C%E2%80%9D+Communication)
18. M. O. E. Elfahal et al., “Automatic recognition and identification for mixed sudanese arabic–english languages speech,” Ph.D. dissertation, Sudan University of Science & Technology, 2019. [find](https://scholar.google.com/scholar?q=M%2E+O%2E+E%2E+Elfahal+et+al%2E%2C+%E2%80%9CAutomatic+recognition+and+identification+for+mixed+sudanese+arabic%E2%80%93english+languages+speech%2C%E2%80%9D+Ph%2ED%2E+dissertation%2C+Sudan+University+of+Science+%26+Technology%2C+2019%2E)
19. I. Hamed, N. T. Vu, and S. Abdennadher, “Arzen: A speech corpus for code-switched egyptian arabic-english,” in Proceedings of The 12th Language Resources and Evaluation Conference, 2020, pp. 4237–4246. [find](https://scholar.google.com/scholar?q=I%2E+Hamed%2C+N%2E+T%2E+Vu%2C+and+S%2E+Abdennadher%2C+%E2%80%9CArzen%3A+A+speech+corpus+for+code%2Dswitched+egyptian+arabic%2Denglish%2C%E2%80%9D+in+Proceedings+of+The+12th+Language+Resources+and+Evaluation+Conference%2C+2020%2C+pp%2E+4237%E2%80%934246)
20. S. A. Chowdhury, A. Hussein, A. Abdelali, and A. Ali, “Towards one model to rule all: Multilingual strategy for dialectal code-switching Arabic Asr,” Proc. Interspeech 2021, 2021. [find](https://scholar.google.com/scholar?q=S%2E+A%2E+Chowdhury%2C+A%2E+Hussein%2C+A%2E+Abdelali%2C+and+A%2E+Ali%2C+%E2%80%9CTowards+one+model+to+rule+all%3A+Multilingual+strategy+for+dialectal+code%2Dswitching+Arabic+Asr%2C%E2%80%9D+Proc%2E+Interspeech+2021%2C+2021%2E) [our card](https://md.research.elamur.ai/free/en/papers/chowdhury-2021-towards-one-model-rule-all/)
21. H. Mubarak, A. Hussein, S. A. Chowdhury, and A. Ali, “Qasr: Qcri aljazeera speech resource a large scale annotated arabic speech corpus,” in ACL, 2021. [find](https://scholar.google.com/scholar?q=H%2E+Mubarak%2C+A%2E+Hussein%2C+S%2E+A%2E+Chowdhury%2C+and+A%2E+Ali%2C+%E2%80%9CQasr%3A+Qcri+aljazeera+speech+resource+a+large+scale+annotated+arabic+speech+corpus%2C%E2%80%9D+in+ACL%2C+2021%2E)
22. M. Mohri, “Weighted automata algorithms,” in Handbook of weighted automata, 2009. [find](https://scholar.google.com/scholar?q=M%2E+Mohri%2C+%E2%80%9CWeighted+automata+algorithms%2C%E2%80%9D+in+Handbook+of+weighted+automata%2C+2009%2E)
23. 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,” in SLT, 2016. [find](https://scholar.google.com/scholar?q=A%2E+Ali%2C+P%2E+Bell%2C+J%2E+Glass%2C+Y%2E+Messaoui%2C+H%2E+Mubarak%2C+S%2E+Renals%2C+and+Y%2E+Zhang%2C+%E2%80%9CThe+MGB%2D2+challenge%3A+Arabic+multi%2Ddialect+broadcast+media+recognition%2C%E2%80%9D+in+SLT%2C+2016%2E) [our card](https://md.research.elamur.ai/free/en/papers/ali-2016-mgb-2-challenge-arabic-multi/)
24. F. Hernandez, V. Nguyen, S. Ghannay, N. Tomashenko, and Y. Estève, “TED-LIUM 3: twice as much data and corpus repartition for experiments on speaker adaptation,” in International Conference on Speech and Computer. Springer, 2018, pp. 198–208. [find](https://scholar.google.com/scholar?q=F%2E+Hernandez%2C+V%2E+Nguyen%2C+S%2E+Ghannay%2C+N%2E+Tomashenko%2C+and+Y%2E+Este%CC%80ve%2C+%E2%80%9CTED%2DLIUM+3%3A+twice+as+much+data+and+corpus+repartition+for+experiments+on+speaker+adaptation%2C%E2%80%9D+in+International+Conference+on+Speech)
25. V. Peddinti, D. Povey, and S. Khudanpur, “A time delay neural network architecture for efficient modeling of long temporal contexts,” in INTERSPEECH, 2015. [find](https://scholar.google.com/scholar?q=V%2E+Peddinti%2C+D%2E+Povey%2C+and+S%2E+Khudanpur%2C+%E2%80%9CA+time+delay+neural+network+architecture+for+efficient+modeling+of+long+temporal+contexts%2C%E2%80%9D+in+INTERSPEECH%2C+2015%2E)
26. D. Povey, V. Peddinti, D. Galvez, P. Ghahrmani, V. Manohar, X. Na, Y. Wang, and S. Khudanpur, “Purely sequence-trained neural networks for ASR based on lattice-free MMI,” in Interspeech, 2016. [find](https://scholar.google.com/scholar?q=D%2E+Povey%2C+V%2E+Peddinti%2C+D%2E+Galvez%2C+P%2E+Ghahrmani%2C+V%2E+Manohar%2C+X%2E+Na%2C+Y%2E+Wang%2C+and+S%2E+Khudanpur%2C+%E2%80%9CPurely+sequence%2Dtrained+neural+networks+for+ASR+based+on+lattice%2Dfree+MMI%2C%E2%80%9D+in+Interspeech%2C+2016%2E)
27. S. Kim and M. L. Seltzer, “Towards language-universal end-to-end speech recognition,” in ICASSP, 2018. [find](https://scholar.google.com/scholar?q=S%2E+Kim+and+M%2E+L%2E+Seltzer%2C+%E2%80%9CTowards+language%2Duniversal+end%2Dto%2Dend+speech+recognition%2C%E2%80%9D+in+ICASSP%2C+2018%2E)
28. A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” in Advances in neural information processing systems, 2017, pp. 5998–6008. [find](https://scholar.google.com/scholar?q=A%2E+Vaswani%2C+N%2E+Shazeer%2C+N%2E+Parmar%2C+J%2E+Uszkoreit%2C+L%2E+Jones%2C+A%2E+N%2E+Gomez%2C+%C5%81%2E+Kaiser%2C+and+I%2E+Polosukhin%2C+%E2%80%9CAttention+is+all+you+need%2C%E2%80%9D+in+Advances+in+neural+information+processing+systems%2C+2017%2C+pp%2E+5998%E2%80%93)
29. T. Kudo and J. Richardson, “Sentencepiece: A simple and language independent subword tokenizer and detokenizer for neural text processing,” arXiv preprint arXiv:1808.06226, 2018. [↗](https://arxiv.org/abs/1808.06226) [our card](https://md.research.elamur.ai/free/en/papers/kudo-2018-sentencepiece-simple-language-independent-subword/)
30. A. Graves, S. Fernández, F. Gomez, and J. Schmidhuber, “Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks,” in Proceedings of the 23rd international conference on Machine learning, 2006. [find](https://scholar.google.com/scholar?q=A%2E+Graves%2C+S%2E+Ferna%CC%81ndez%2C+F%2E+Gomez%2C+and+J%2E+Schmidhuber%2C+%E2%80%9CConnectionist+temporal+classification%3A+labelling+unsegmented+sequence+data+with+recurrent+neural+networks%2C%E2%80%9D+in+Proceedings+of+the+23rd+internat)
31. M. Mohri, F. Pereira, and M. Riley, “Speech recognition with weighted finite-state transducers,” in Springer Handbook of Speech Processing, 2008. [find](https://scholar.google.com/scholar?q=M%2E+Mohri%2C+F%2E+Pereira%2C+and+M%2E+Riley%2C+%E2%80%9CSpeech+recognition+with+weighted+finite%2Dstate+transducers%2C%E2%80%9D+in+Springer+Handbook+of+Speech+Processing%2C+2008%2E)
32. D. Povey, A. Ghoshal, G. Boulianne, L. Burget, O. Glembek, N. Goel, M. Hannemann, P. Motlicek, Y. Qian, P. Schwarz et al., “The Kaldi speech recognition toolkit,” in ASRU, 2011. [find](https://scholar.google.com/scholar?q=D%2E+Povey%2C+A%2E+Ghoshal%2C+G%2E+Boulianne%2C+L%2E+Burget%2C+O%2E+Glembek%2C+N%2E+Goel%2C+M%2E+Hannemann%2C+P%2E+Motlicek%2C+Y%2E+Qian%2C+P%2E+Schwarz+et+al%2E%2C+%E2%80%9CThe+Kaldi+speech+recognition+toolkit%2C%E2%80%9D+in+ASRU%2C+2011%2E)
33. G. D. Forney, “The Viterbi algorithm,” Proceedings of the IEEE, vol. 61, no. 3, pp. 268–278, 1973. [find](https://scholar.google.com/scholar?q=G%2E+D%2E+Forney%2C+%E2%80%9CThe+Viterbi+algorithm%2C%E2%80%9D+Proceedings+of+the+IEEE%2C+vol%2E+61%2C+no%2E+3%2C+pp%2E+268%E2%80%93278%2C+1973%2E)
34. C. Liu, Q. Zhang, X. Zhang, K. Singh, Y. Saraf, and G. Zweig, “Multilingual ASR with massive data augmentation,” arXiv preprint arXiv:1909.06522, 2019. [↗](https://arxiv.org/abs/1909.06522)
35. E. Yılmaz, S. Cohen, X. Yue, D. van Leeuwen, and H. Li, “Multi-graph decoding for code-switching ASR,” arXiv preprint arXiv:1906.07523, 2019. [↗](https://arxiv.org/abs/1906.07523)
36. S. A. Chowdhury, Y. Samih, M. Eldesouki, and A. Ali, “Effects of dialectal code-switching on speech modules: A study using egyptian Arabic broadcast speech,” Proc. Interspeech, 2020. [find](https://scholar.google.com/scholar?q=S%2E+A%2E+Chowdhury%2C+Y%2E+Samih%2C+M%2E+Eldesouki%2C+and+A%2E+Ali%2C+%E2%80%9CEffects+of+dialectal+code%2Dswitching+on+speech+modules%3A+A+study+using+egyptian+Arabic+broadcast+speech%2C%E2%80%9D+Proc%2E+Interspeech%2C+2020%2E)