---
title: "Romanized Arabic Across Dialects — Keleg 2026"
description: "Most Arabic NLP work treats Arabizi (Arabic written in Latin letters and digits) as a leftover from the era of poor Arabic-script support."
authors: ["Amr Keleg", "Ahmed Amine Ben Abdallah", "Taha Yassine", "Chadi Helwe", "Imane Guellil", "Nedjma Ousidhoum"]
year: 2026
venue: "arXiv:2608.02555"
source: "https://arxiv.org/abs/2608.02555"
doi: "10.48550/arxiv.2608.02555"
arxiv: "2608.02555"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/keleg-2026-romanized-arabic-across-dialects-views/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/keleg-2026-romanized-arabic-across-dialects-views/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/) / Romanized Arabic Across Dialects: Views, Usage Patterns, and Linguistic Variation

# Romanized Arabic Across Dialects: Views, Usage Patterns, and Linguistic Variation

**Authors:** [Amr Keleg](https://md.research.elamur.ai/free/en/authors/amr-keleg/), [Ahmed Amine Ben Abdallah](https://md.research.elamur.ai/free/en/authors/ahmed-amine-ben-abdallah/), Taha Yassine, [Chadi Helwe](https://md.research.elamur.ai/free/en/authors/chadi-helwe/), [Imane Guellil](https://md.research.elamur.ai/free/en/authors/imane-guellil/), [Nedjma Ousidhoum](https://md.research.elamur.ai/free/en/authors/nedjma-ousidhoum/) · **Year:** 2026 · **Venue:** arXiv:2608.02555 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2608.02555) · [PDF](https://arxiv.org/pdf/2608.02555) · [DOI 10.48550/arxiv.2608.02555](https://doi.org/10.48550/arxiv.2608.02555) · [arXiv 2608.02555](https://arxiv.org/abs/2608.02555) · [OpenAlex](https://openalex.org/W7172453400)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14)

## Summary
Most Arabic NLP work treats Arabizi (Arabic written in Latin letters and digits) as a leftover from the era of poor Arabic-script support. It also tends to study one country at a time. This paper challenges that view with a two-part online survey of 189 Arabic speakers. The first part asks how participants perceive and use Arabizi. The second part has them transliterate Arabic words and sentences.

The analysis centres on five national dialects: Algerian, Moroccan and Tunisian from the Maghreb, and Egyptian and Lebanese from the Mashriq. The authors build two resources:
- a character-aligned lexicon of participant transliterations;
- a parallel corpus pairing Arabic-script sentences with several Arabizi versions.

Attitudes differ across countries. Use appears to be declining in Egypt, yet some speakers still rely on Arabizi heavily. Within a dialect, spelling choices are fairly regular, while certain letters are rendered differently across dialects. Native speakers can almost always spot transliterations written by people from their own country, even without lexical clues. The work matters for building dialect-aware tools, because it shows Arabizi is a living, patterned writing practice rather than noise.

## Key points
- The survey ran on Qualtrics from June 9 to August 31, 2025, in Arabic, English and French. It used snowball sampling through X, LinkedIn, personal networks and community lists (SIGARAB, Masakhane, North Africans in NLP). Of 189 responses, 170 completed both parts, with a median time of about 11 minutes. 144 responses from the five focus countries were analysed.
- Perceptions follow two patterns:
  - Lebanon, Morocco and Tunisia show a split, with some saying Arabizi has limited use and others preferring it.
  - Algeria and Egypt lean toward "limited contexts": 13 of 21 Algerians and 22 of 33 Egyptians.
  - Six Egyptians and two Algerians say Arabizi should never be used.
- The lexicon covers 210 Arabic words (at least 40 per dialect), averaging 20.8 transliterations per word, with character-level alignment.
- The parallel corpus has 189 Arabic–Arabizi sentences, chosen to avoid country-specific vocabulary. Dialect identification must therefore rest on spelling and phonological cues.
- Annotators identified same-country transliterations almost perfectly and generally recognised cues from other dialects.

## Interesting ideas
- **Many writers per item.** Collecting spellings of one fixed word list from many writers lets you measure variation within a group separately from variation between groups. — For Mova (tokens): A similar crowd-style variant list for English (regional, informal or phonetic spellings), mapped to one lemma, could feed an explainable normalisation table in front of the tagger.
- **Character-level alignment of variants.** Aligning each variant with its source word letter by letter shows exactly which positions vary, instead of only counting how many variants a word has. — For Mova (tokens): Aligning nonstandard spellings to canonical forms could yield transparent per-character rewrite rules, rather than a black-box spelling corrector.
- **Stimuli with lexical markers removed.** Sentences without telltale vocabulary force judges to rely only on the cue under study, which works as a built-in control against shortcut signals. — For Mova (learn): Evaluation sets for dialect conversion or genre classifiers could strip giveaway words, testing whether a model uses the intended structural signal rather than keyword shortcuts.
- **"Not sure" as a valid answer.** The identification task allowed "my country", "another country" or "not sure", so uncertainty was recorded instead of being forced into a guess. — For Mova (answer): Readers answering fable or school-test questions could have an explicit abstain option, scored separately, so that confident errors stand apart from honest uncertainty.

*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/)

## Cite

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

```bibtex
@misc{keleg2026romanized,
  title = {{Romanized Arabic Across Dialects: Views, Usage Patterns, and Linguistic Variation}},
  author = {Keleg, Amr and Abdallah, Ahmed Amine Ben and Yassine, Taha and Helwe, Chadi and Guellil, Imane and Ousidhoum, Nedjma},
  year = {2026},
  doi = {10.48550/arxiv.2608.02555},
  eprint = {2608.02555},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2608.02555}
}
```

- [← previous Roles of Predictability and Acoustic Distance in Sound Discrimination via Contrastive Learning](https://md.research.elamur.ai/free/en/papers/zhang-2026-roles-predictability-acoustic-distance-sound/)
- [next → Rosetta at AlexandriaX-2026: LoRA-Adapted NileChat for Context-Aware Dialectal Arabic Dialogue Translation](https://md.research.elamur.ai/free/en/papers/esmaeil-2026-rosetta-alexandriax-2026-lora-adapted/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/)

## References

22 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Muhammad Abdul-Mageed, Amr Keleg, Abdel-Rahim Elmadany, Chiyu Zhang, Injy Hamed, Walid Magdy, Houda Bouamor, and Nizar Habash. 2024. NADI 2024: The fifth nuanced Arabic dialect identification shared task. In Proceedings of the Second Arabic Natural Language Processing Conference, pages 709--728, Bangkok, Thailand. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Muhammad+Abdul%2DMageed%2C+Amr+Keleg%2C+Abdel%2DRahim+Elmadany%2C+Chiyu+Zhang%2C+Injy+Hamed%2C+Walid+Magdy%2C+Houda+Bouamor%2C+and+Nizar+Habash%2E+2024%2E+NADI+2024%3A+The+fifth+nuanced+Arabic+dialect+identification+shared+t)
2. Verena Blaschke, Christoph Purschke, Hinrich Schuetze, and Barbara Plank. 2024. What do dialect speakers want? a survey of attitudes towards language technology for German dialects. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 823--841, Bangkok, Thailand. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Verena+Blaschke%2C+Christoph+Purschke%2C+Hinrich+Schuetze%2C+and+Barbara+Plank%2E+2024%2E+What+do+dialect+speakers+want%3F+a+survey+of+attitudes+towards+language+technology+for+German+dialects%2E+In+Proceedings+of+)
3. Aula Khatteb Abu-Liel, Zohar Eviatar, and Bracha Nir. 2019. Writing between languages: The case of Arabizi. Writing Systems Research, 11(2):226--238. [find](https://scholar.google.com/scholar?q=Aula+Khatteb+Abu%2DLiel%2C+Zohar+Eviatar%2C+and+Bracha+Nir%2E+2019%2E+Writing+between+languages%3A+The+case+of+Arabizi%2E+Writing+Systems+Research%2C+11%282%29%3A226%2D%2D238%2E)
4. Wafia Adouane, Nasredine Semmar, and Richard Johansson. 2016. Romanized Berber and Romanized Arabic automatic language identification using machine learning. In Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial3), Tim Buckwalter. 2004. Buckwalter Arabic morphological analyzer version 2.0 (ldc2004l02). Web Download. LDC catalog number LDC2004L02, ISBN 1-58563-324-0. [find](https://scholar.google.com/scholar?q=Wafia+Adouane%2C+Nasredine+Semmar%2C+and+Richard+Johansson%2E+2016%2E+Romanized+Berber+and+Romanized+Arabic+automatic+language+identification+using+machine+learning%2E+In+Proceedings+of+the+Third+Workshop+on+NL)
5. Anna Kashina. 2020. Case study of language preferences in social media of Tunisia. In Proceedings of the International Conference Digital Age: Traditions, Modernity and Innovations (ICDATMI 2020), pages 111--115. Atlantis Press. [find](https://scholar.google.com/scholar?q=Anna+Kashina%2E+2020%2E+Case+study+of+language+preferences+in+social+media+of+Tunisia%2E+In+Proceedings+of+the+International+Conference+Digital+Age%3A+Traditions%2C+Modernity+and+Innovations+%28ICDATMI+2020%29%2C+pag)
6. Ryan Cotterell, Adithya Renduchintala, Naomi Saphra, and Chris Callison-Burch. 2014. An Algerian Arabic-French code-switched corpus. In Workshop on free/open-source Arabic corpora and corpora processing tools, page 34. [find](https://scholar.google.com/scholar?q=Ryan+Cotterell%2C+Adithya+Renduchintala%2C+Naomi+Saphra%2C+and+Chris+Callison%2DBurch%2E+2014%2E+An+Algerian+Arabic%2DFrench+code%2Dswitched+corpus%2E+In+Workshop+on+free%2Fopen%2Dsource+Arabic+corpora+and+corpora+processi)
7. Amr Keleg. 2025. LLM alignment for the Arabs: A homogenous culture or diverse ones. In Proceedings of the 3rd Workshop on Cross-Cultural Considerations in NLP (C3NLP 2025), pages 1--9, Albuquerque, New Mexico. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Amr+Keleg%2E+2025%2E+LLM+alignment+for+the+Arabs%3A+A+homogenous+culture+or+diverse+ones%2E+In+Proceedings+of+the+3rd+Workshop+on+Cross%2DCultural+Considerations+in+NLP+%28C3NLP+2025%29%2C+pages+1%2D%2D9%2C+Albuquerque%2C+Ne)
8. Kareem Darwish. 2014. Arabizi detection and conversion to Arabic. In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP), pages 217--224, Doha, Qatar. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Kareem+Darwish%2E+2014%2E+Arabizi+detection+and+conversion+to+Arabic%2E+In+Proceedings+of+the+EMNLP+2014+Workshop+on+Arabic+Natural+Language+Processing+%28ANLP%29%2C+pages+217%2D%2D224%2C+Doha%2C+Qatar%2E+Association+for+C)
9. Amr Keleg, Sharon Goldwater, and Walid Magdy. 2023. ALDi: Quantifying the Arabic level of dialectness of text. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 10597--10611, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Amr+Keleg%2C+Sharon+Goldwater%2C+and+Walid+Magdy%2E+2023%2E+ALDi%3A+Quantifying+the+Arabic+level+of+dialectness+of+text%2E+In+Proceedings+of+the+2023+Conference+on+Empirical+Methods+in+Natural+Language+Processing)
10. Ronald A. Fisher. 1935. The Principles of Experimentation, Illustrated by a Psycho-physical Experiment, chapter 2. Oliver and Boyd, Edinburgh. [find](https://scholar.google.com/scholar?q=Ronald+A%2E+Fisher%2E+1935%2E+The+Principles+of+Experimentation%2C+Illustrated+by+a+Psycho%2Dphysical+Experiment%2C+chapter+2%2E+Oliver+and+Boyd%2C+Edinburgh%2E)
11. Chayma Fourati, Hatem Haddad, Abir Messaoudi, Moez BenHajhmida, Aymen Ben El-haj Mabrouk, and Malek Naski. 2021. Introducing a large Tunisian Arabizi dialectal dataset for sentiment analysis. In Proceedings of the Sixth Arabic Natural Language Processing Workshop, pages 226--230, Kyiv, Ukraine (Virtual). Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Chayma+Fourati%2C+Hatem+Haddad%2C+Abir+Messaoudi%2C+Moez+BenHajhmida%2C+Aymen+Ben+El%2Dhaj+Mabrouk%2C+and+Malek+Naski%2E+2021%2E+Introducing+a+large+Tunisian+Arabizi+dialectal+dataset+for+sentiment+analysis%2E+In+Proce)
12. Amr Keleg, Sharon Goldwater, and Walid Magdy. 2025. Revisiting common assumptions about Arabic dialects in NLP. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3309--3327, Vienna, Austria. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Amr+Keleg%2C+Sharon+Goldwater%2C+and+Walid+Magdy%2E+2025%2E+Revisiting+common+assumptions+about+Arabic+dialects+in+NLP%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+the+Association+for+Computational+Linguisti)
13. Yves Gonzalez-Quijano. 2014. Technology Literacies of the New Media: Phrasing the World in the “Arab Easy” (R)evolution, pages 159--166. Palgrave Macmillan US, New York. [find](https://scholar.google.com/scholar?q=Yves+Gonzalez%2DQuijano%2E+2014%2E+Technology+Literacies+of+the+New+Media%3A+Phrasing+the+World+in+the+%E2%80%9CArab+Easy%E2%80%9D+%28R%29evolution%2C+pages+159%2D%2D166%2E+Palgrave+Macmillan+US%2C+New+York%2E)
14. Karen McNeil. 2022. we don’t speak the same language: language choice and identity on a Tunisian internet forum. International Journal of the Sociology of Language, 2022(278):51--80. [find](https://scholar.google.com/scholar?q=Karen+McNeil%2E+2022%2E+we+don%E2%80%99t+speak+the+same+language%3A+language+choice+and+identity+on+a+Tunisian+internet+forum%2E+International+Journal+of+the+Sociology+of+Language%2C+2022%28278%29%3A51%2D%2D80%2E)
15. Imane Guellil, Faical Azouaou, Fodil Benali, Ala Eddine Hachani, and Marcelo Mendoza. 2020. The Role of Transliteration in the Process of Arabizi Translation/Sentiment Analysis, pages 101--128. Springer International Publishing, Cham. [find](https://scholar.google.com/scholar?q=Imane+Guellil%2C+Faical+Azouaou%2C+Fodil+Benali%2C+Ala+Eddine+Hachani%2C+and+Marcelo+Mendoza%2E+2020%2E+The+Role+of+Transliteration+in+the+Process+of+Arabizi+Translation%2FSentiment+Analysis%2C+pages+101%2D%2D128%2E+Spring)
16. Leila Moudjari, Karima Akli-Astouati, and Farah Benamara. 2020. An Algerian corpus and an annotation platform for opinion and emotion analysis. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 1202--1210, Marseille, France. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Leila+Moudjari%2C+Karima+Akli%2DAstouati%2C+and+Farah+Benamara%2E+2020%2E+An+Algerian+corpus+and+an+annotation+platform+for+opinion+and+emotion+analysis%2E+In+Proceedings+of+the+Twelfth+Language+Resources+and+Eva)
17. Nizar Y Habash. 2010. Introduction to Arabic Natural Language Processing. Morgan & Claypool Publishers. [find](https://scholar.google.com/scholar?q=Nizar+Y+Habash%2E+2010%2E+Introduction+to+Arabic+Natural+Language+Processing%2E+Morgan+%26+Claypool+Publishers%2E)
18. Arij Riabi, Menel Mahamdi, and Djamé Seddah. 2023. Enriching the NArabizi treebank: A multifaceted approach to supporting an under-resourced language. In Proceedings of the 17th Linguistic Annotation Workshop (LAW-XVII), pages 266--278, Toronto, Canada. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Arij+Riabi%2C+Menel+Mahamdi%2C+and+Djam%C3%A9+Seddah%2E+2023%2E+Enriching+the+NArabizi+treebank%3A+A+multifaceted+approach+to+supporting+an+under%2Dresourced+language%2E+In+Proceedings+of+the+17th+Linguistic+Annotation+)
19. Karen Holtzblatt and Hugh Beyer. 2017. The affinity diagram. In Contextual Design, second edition edition, Interactive Technologies, pages 127--146. Morgan Kaufmann, Boston. [find](https://scholar.google.com/scholar?q=Karen+Holtzblatt+and+Hugh+Beyer%2E+2017%2E+The+affinity+diagram%2E+In+Contextual+Design%2C+second+edition+edition%2C+Interactive+Technologies%2C+pages+127%2D%2D146%2E+Morgan+Kaufmann%2C+Boston%2E)
20. Samia A Jaran and Fawwaz Al-Abed Al-Haq. 2015. The use of hybrid terms and expressions in Colloquial Arabic among Jordanian college students: A sociolinguistic study. English Language Teaching, 8(12):86--97. [find](https://scholar.google.com/scholar?q=Samia+A+Jaran+and+Fawwaz+Al%2DAbed+Al%2DHaq%2E+2015%2E+The+use+of+hybrid+terms+and+expressions+in+Colloquial+Arabic+among+Jordanian+college+students%3A+A+sociolinguistic+study%2E+English+Language+Teaching%2C+8%2812%29%3A)
21. Houda Saadane and Nizar Habash. 2015. A conventional orthography for Algerian Arabic. In Proceedings of the Second Workshop on Arabic Natural Language Processing, pages 69--79, Beijing, China. Association for Computational Linguistics. Other represented dialects are: Jordan (16), Saudi Arabia (13), Palestine (8), United Arab Emirates (7), Kuwait (5), Oman (4), and Libya (1). However, most of the manuscripts that consider these dialects are sociolinguistic studies, and the few that released NLP datasets were not published in *CL conferences/workshops. [find](https://scholar.google.com/scholar?q=Houda+Saadane+and+Nizar+Habash%2E+2015%2E+A+conventional+orthography+for+Algerian+Arabic%2E+In+Proceedings+of+the+Second+Workshop+on+Arabic+Natural+Language+Processing%2C+pages+69%2D%2D79%2C+Beijing%2C+China%2E+Associa)
22. Wael Salloum. 2018. Machine Translation of Arabic Dialects. PhD thesis, Columbia University. B Analysis of the Users’ Preferred Scripts Taha Tobaili. 2016. Arabizi identification in Twitter data. In Proceedings of the ACL 2016 Student Research Workshop, pages 51--57, Berlin, Germany. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Wael+Salloum%2E+2018%2E+Machine+Translation+of+Arabic+Dialects%2E+PhD+thesis%2C+Columbia+University%2E+B+Analysis+of+the+Users%E2%80%99+Preferred+Scripts+Taha+Tobaili%2E+2016%2E+Arabizi+identification+in+Twitter+data%2E+In+P) [our card](https://md.research.elamur.ai/free/en/papers/zbib-2012-machine-translation-arabic-dialects/)