---
title: "Classifying Arabic dialect text in the Social Media Arabic Dialect"
description: "Arabic speakers increasingly write in regional dialects on social media, but automatic dialect identification has… 88% accuracy. 19 references."
authors: ["Areej Alshutayri", "Eric Atwell"]
year: 2019
venue: "Proceedings of the 3rd Workshop on Arabic Corpus Linguistics"
source: "https://aclanthology.org/W19-5607/"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/alshutayri-2019-classifying-arabic-dialect-text-social/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/alshutayri-2019-classifying-arabic-dialect-text-social/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2019](https://md.research.elamur.ai/free/en/papers/2019/page/7/) / Classifying Arabic dialect text in the Social Media Arabic Dialect Corpus (SMADC)

# Classifying Arabic dialect text in the Social Media Arabic Dialect Corpus (SMADC)

**Authors:** [Areej Alshutayri](https://md.research.elamur.ai/free/en/authors/areej-alshutayri/) and [Eric Atwell](https://md.research.elamur.ai/free/en/authors/eric-atwell/) · **Year:** 2019 · **Venue:** [Proceedings of the 3rd Workshop on Arabic Corpus Linguistics](https://md.research.elamur.ai/free/en/venues/wacl/) ([Workshop on Arabic Corpus Linguistics 2019](https://md.research.elamur.ai/free/en/venues/wacl/2019/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/W19-5607/) · [PDF](https://aclanthology.org/W19-5607.pdf) · [OpenAlex](https://openalex.org/W3088428493)

**Citations:** 5 citations (OpenAlex, 2026-10-14)

## Summary
Arabic speakers increasingly write in regional dialects on social media, but automatic dialect identification has lagged because few labelled dialect corpora exist. The authors address this with the Social Media Arabic Dialect Corpus (SMADC), built from tweets, Facebook posts and comments on online newspapers. It covers the five main regional groups: Gulf, Iraqi, Egyptian, Levantine and North African. Crowd annotators labelled each text and also marked the words that revealed its dialect. The classifier is purely lexicon-based. It first strips Modern Standard Arabic (MSA) words using a large stop list, then matches the remaining words against per-dialect dictionaries built from the training documents. Several scoring schemes are compared: matching against annotator-supplied cue terms, simple and weighted voting, and frequency-based weights combined either by averaging (WAM) or by multiplication (WMM). The averaged frequency weights work best, reaching 88% accuracy, and 90% once the MSA list is cleaned of mislabelled dialect words. This shows that a transparent, dictionary-driven approach can be competitive when the lexicons come from balanced, annotated data.

## Key points
- SMADC covers EGY, GLF, LEV, IRQ and NOR. The MSA filter list, called StopWords1, merges corpus MSA documents, web stop words, a Sketch Engine list and translated seed words, for 29,674 entries in total.
- The evaluation subset has 1633 documents: 1383 for training (18,697 tokens) and 250 held out for testing (7,341 tokens).
- **Dialectal terms method:**
  - The annotator cue-word dictionaries are small, with 312–451 words per dialect.
  - The best accuracy is 56.91%, with 110 test documents left unclassified because shared words matched more than one dialect.
- **Voting:** weighted voting reaches 74%, rising to 77.60% after the MSA list was cleaned.
- **Frequency-weight methods:**
  - WAM scores 88% with MSA removal and 64% without it, improving to 90% after cleaning.
  - WMM scores 55.60% with MSA removal and 43.0% without it.

## Interesting ideas
- Removing everything that belongs to the shared standard variety before scoring leaves only the distinctive residue, which sharpened classification considerably (88% versus 64%).
- The system outputs an explicit "unclassified" verdict when evidence ties across classes, instead of guessing, so ambiguity stays visible.
- Combining per-word weights by length-normalised averaging proved more robust than multiplying them, because a product is easily dominated or collapsed by a single missing or extreme item.
- For words absent from a dictionary, the authors substitute one divided by that dictionary's size instead of zero or one, so the penalty scales with how large each lexicon is.
- Annotators labelled each text and also wrote down the specific words that justified their label, producing ready-made explanatory lexicons.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (tokens): Before choosing a register- or variety-specific lexicon, the tagger could set aside common-core English words and decide on the variety using only the leftover marked vocabulary.
- **I2** (learn): When witness parsers tie or split evenly, the sentence should be flagged as abstained rather than admitted as silver data, and the abstention rate should be tracked as its own metric.
- **I3** (frames): When scoring candidate VerbNet classes or frames from several cue words, an averaged evidence score could avoid one absent cue vetoing an otherwise well-supported reading.
- **I4** (tokens): Unknown-word backoff in per-domain or per-variety lexicons could be scaled by lexicon size, keeping comparisons fair between big and small resources.
- **I5** (learn): LLM teacher tiers could be asked to name the tokens supporting each label, yielding auditable evidence and seed dictionaries for self-induced rules.

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

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (1): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works

## Cite

Cite as: Areej Alshutayri and Eric Atwell. 2019. Classifying Arabic dialect text in the Social Media Arabic Dialect Corpus (SMADC). Proceedings of the 3rd Workshop on Arabic Corpus Linguistics.

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

```bibtex
@inproceedings{alshutayri2019classifying,
  title = {{Classifying Arabic dialect text in the Social Media Arabic Dialect Corpus (SMADC)}},
  author = {Areej Alshutayri and Eric Atwell},
  booktitle = {Proceedings of the 3rd Workshop on Arabic Corpus Linguistics},
  year = {2019},
  url = {https://aclanthology.org/W19-5607/}
}
```

- [← previous Application of an Automatic Plagiarism Detection System in a Large-scale Assessment of English Speaking Proficiency](https://md.research.elamur.ai/free/en/papers/wang-2019-application-automatic-plagiarism-detection-system/)
- [next → Comparing Pipelined and Integrated Approaches to Dialectal Arabic Neural Machine Translation](https://md.research.elamur.ai/free/en/papers/shapiro-2019-comparing-pipelined-integrated-approaches-dialectal/)
- [↑ back to papers of 2019](https://md.research.elamur.ai/free/en/papers/2019/page/7/)

## References

19 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Wafia Adouane and Simon Dobnik. 2017. [Identification of languages in Algerian Arabic multilingual documents](https://md.research.elamur.ai/free/en/papers/adouane-2017-identification-languages-algerian-arabic-multilingual/). In Proceedings of the Third Arabic Natural Language Processing Workshop, pages 1– 8, Valencia, Spain. Association for Computational Linguistics.
2. Mona Al-Walaie and Muhammad Khan. 2017. Arabic dialects classification using text mining techniques. In 2017 International Conference on Computer and Applications (ICCA), pages 325–329. [find](https://scholar.google.com/scholar?q=Mona+Al%2DWalaie+and+Muhammad+Khan%2E+2017%2E+Arabic+dialects+classification+using+text+mining+techniques%2E+In+2017+International+Conference+on+Computer+and+Applications+%28ICCA%29%2C+pages+325%E2%80%93329%2E)
3. Areej Alshutayri and Eric Atwell. 2017. Exploring twitter as a source of an arabic dialect corpus. International Journal of Computational Linguistics (IJCL), 8:37–44. [find](https://scholar.google.com/scholar?q=Areej+Alshutayri+and+Eric+Atwell%2E+2017%2E+Exploring+twitter+as+a+source+of+an+arabic+dialect+corpus%2E+International+Journal+of+Computational+Linguistics+%28IJCL%29%2C+8%3A37%E2%80%9344%2E)
4. Areej Alshutayri and Eric Atwell. 2018a. Arabic dialects annotation using an online game. In 2nd International Conference on Natural Language and Speech Processing (ICNLSP), pages 1–5. [find](https://scholar.google.com/scholar?q=Areej+Alshutayri+and+Eric+Atwell%2E+2018a%2E+Arabic+dialects+annotation+using+an+online+game%2E+In+2nd+International+Conference+on+Natural+Language+and+Speech+Processing+%28ICNLSP%29%2C+pages+1%E2%80%935%2E)
5. Areej Alshutayri and Eric Atwell. 2018b. Creating an arabic dialect text corpus by exploring twitter, facebook, and online newspapers. In Proceedings of Open-Source Arabic Corpora and Processing Tools. OSACT’2018 Open-Source, Miyazaki, Japan. [find](https://scholar.google.com/scholar?q=Areej+Alshutayri+and+Eric+Atwell%2E+2018b%2E+Creating+an+arabic+dialect+text+corpus+by+exploring+twitter%2C+facebook%2C+and+online+newspapers%2E+In+Proceedings+of+Open%2DSource+Arabic+Corpora+and+Processing+Tools)
6. Areej Alshutayri and Eric Atwell. 2018c. A Social Media Corpus of Arabic Dialect Text. Computer-Mediated Communication and Social Media Corpora, Clermont-Ferrand: Presses universitaires Blaise Pascal. [find](https://scholar.google.com/scholar?q=Areej+Alshutayri+and+Eric+Atwell%2E+2018c%2E+A+Social+Media+Corpus+of+Arabic+Dialect+Text%2E+Computer%2DMediated+Communication+and+Social+Media+Corpora%2C+Clermont%2DFerrand%3A+Presses+universitaires+Blaise+Pascal%2E)
7. Heba Elfardy and Mona Diab. 2013. [Sentence level dialect identification in Arabic](https://md.research.elamur.ai/free/en/papers/elfardy-2013-sentence-level-dialect-identification-arabic/). In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, volume 2, pages 456–461, Sofia, Bulgaria. Association for Computational Linguistics.
8. Imène Guellil and Faiçal Azouaou. 2016. Arabic dialect identification with an unsupervised learning (based on a lexicon). application case: Algerian dialect. In IEEE Intl Conference on Computational Science and Engineering (CSE) and IEEE Intl Conference on Embedded and Ubiquitous Computing (EUC) and 15th Intl Symposium on Distributed Computing and Applications for Business Engineering (DCABES), pages 724–731. [find](https://scholar.google.com/scholar?q=Ime%CC%80ne+Guellil+and+Faic%CC%A7al+Azouaou%2E+2016%2E+Arabic+dialect+identification+with+an+unsupervised+learning+%28based+on+a+lexicon%29%2E+application+case%3A+Algerian+dialect%2E+In+IEEE+Intl+Conference+on+Computational)
9. Nizar Y. Habash. 2010. Introduction to Arabic Natural Language Processing. Morgan and Claypool. [↗](https://doi.org/10.1007/978-3-031-02139-8)
10. Salima Harrat, Karima Meftouh, Mourad Abbas, and Kamel Smaı̈li. 2014. Building resources for algerian arabic dialects. In 15th Annual Conference of the International Communication Association (Interspeech), pages 2123–2127. [find](https://scholar.google.com/scholar?q=Salima+Harrat%2C+Karima+Meftouh%2C+Mourad+Abbas%2C+and+Kamel+Sma%C4%B1%CC%88li%2E+2014%2E+Building+resources+for+algerian+arabic+dialects%2E+In+15th+Annual+Conference+of+the+International+Communication+Association+%28Intersp)
11. Emmanouil Ikonomakis, Sotiris Kotsiantis, and V Tampakas. 2005. Text classification using machine learning techniques. WSEAS transactions on computers, 4:966–974. [find](https://scholar.google.com/scholar?q=Emmanouil+Ikonomakis%2C+Sotiris+Kotsiantis%2C+and+V+Tampakas%2E+2005%2E+Text+classification+using+machine+learning+techniques%2E+WSEAS+transactions+on+computers%2C+4%3A966%E2%80%93974%2E)
12. Adam Kilgarriff, Vı́t Baisa, Jan Bušta, Miloš Jakubı́ček, Vojtěch Kovář, Jan Michelfeit, Pavel Rychlý, and Vı́t Suchomel. 2014. The sketch engine: ten years on. Lexicography, 1(1):7–36. [find](https://scholar.google.com/scholar?q=Adam+Kilgarriff%2C+V%C4%B1%CC%81t+Baisa%2C+Jan+Bus%CC%8Cta%2C+Milos%CC%8C+Jakub%C4%B1%CC%81c%CC%8Cek%2C+Vojte%CC%8Cch+Kova%CC%81r%CC%8C%2C+Jan+Michelfeit%2C+Pavel+Rychly%CC%81%2C+and+V%C4%B1%CC%81t+Suchomel%2E+2014%2E+The+sketch+engine%3A+ten+years+on%2E+Lexicography%2C+1%281%29%3A7%E2%80%9336%2E)
13. Hamdy Mubarak and Kareem Darwish. 2014. [Using twitter to collect a multi-dialectal corpus of Arabic](https://md.research.elamur.ai/free/en/papers/mubarak-2014-twitter-collect-multi-dialectal-corpus/). In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP), pages 1–7, Doha, Qatar. Association for Computational Linguistics.
14. Arfath Pasha, Mohamed Al-Badrashiny, Mona Diab, Ahmed El Kholy, Ramy Eskander, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. [MADAMIRA: A fast, comprehensive tool for morphological analysis and disambiguation of Arabic](https://md.research.elamur.ai/free/en/papers/pasha-2014-madamira-fast-comprehensive-tool-morphological/). In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 1094–1101, Reykjavik, Iceland. European Language Resources Association (ELRA).
15. Hanna Sababa and Athena Stassopoulou. 2018. A classifier to distinguish between cypriot greek and standard modern greek. pages 251–255. [find](https://scholar.google.com/scholar?q=Hanna+Sababa+and+Athena+Stassopoulou%2E+2018%2E+A+classifier+to+distinguish+between+cypriot+greek+and+standard+modern+greek%2E+pages+251%E2%80%93255%2E)
16. Fatiha Sadat, Farnzeh Kazemi, and Atefeh Farzindar. 2014. [Automatic identification of Arabic language varieties and dialects in social media](https://md.research.elamur.ai/free/en/papers/sadat-2014-automatic-identification-arabic-language-varieties/). In Proceedings of the Second Workshop on Natural Language Processing for Social Media (SocialNLP), pages 22– 27, Dublin, Ireland. Association for Computational Linguistics and Dublin City University.
17. Serge Sharoff. 2006. Open-source corpora: Using the net to fish for linguistic data. International Journal of Corpus Linguistics, 11:435–462. [find](https://scholar.google.com/scholar?q=Serge+Sharoff%2E+2006%2E+Open%2Dsource+corpora%3A+Using+the+net+to+fish+for+linguistic+data%2E+International+Journal+of+Corpus+Linguistics%2C+11%3A435%E2%80%93462%2E)
18. Abdulhadi Shoufan and Sumaya Alameri. 2015. [Natural language processing for dialectical arabic: A survey](https://md.research.elamur.ai/free/en/papers/shoufan-2015-natural-language-processing-dialectical-arabic/). In Proceedings of the Second Workshop on Arabic Natural Language Processing, pages 36–48, Beijing, China. Association for Computational Linguistics.
19. Omar F. Zaidan and Chris Callison-Burch. 2014. [Arabic dialect identification](https://md.research.elamur.ai/free/en/papers/zaidan-2014-arabic-dialect-identification/). Computational Linguistics, 40(1):171–202. [↗](https://doi.org/10.1162/coli_a_00169)

*Updated 1 November 2026*