---
title: "Improved Arabic Dialect Classification with Social Media Data"
description: "The paper tackles the problem of telling apart Modern Standard Arabic (MSA) and regional Arabic… 19 references, cited by 16 papers here."
authors: ["Fei Huang"]
year: 2015
venue: "Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing"
source: "https://aclanthology.org/D15-1254/"
doi: "10.18653/v1/d15-1254"
licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)"
licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/huang-2015-improved-arabic-dialect-classification-social/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/huang-2015-improved-arabic-dialect-classification-social/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2015](https://md.research.elamur.ai/free/en/papers/2015/page/3/) / Improved Arabic Dialect Classification with Social Media Data

# Improved Arabic Dialect Classification with Social Media Data

**Authors:** [Fei Huang](https://md.research.elamur.ai/free/en/authors/fei-huang/) · **Year:** 2015 · **Venue:** [Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing](https://md.research.elamur.ai/free/en/venues/emnlp/) ([EMNLP 2015](https://md.research.elamur.ai/free/en/venues/emnlp/2015/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/)

**Links:** [Publisher's page](https://aclanthology.org/D15-1254/) · [PDF](https://aclanthology.org/D15-1254.pdf) · [DOI 10.18653/v1/d15-1254](https://doi.org/10.18653/v1/d15-1254) · [OpenAlex](https://openalex.org/W2251867433)

**Citations:** 42 citations (OpenAlex, 2026-10-14) · [16](https://md.research.elamur.ai/free/en/papers/huang-2015-improved-arabic-dialect-classification-social/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper tackles the problem of telling apart Modern Standard Arabic (MSA) and regional Arabic dialects in social media posts. This matters because most Arabic-English MT resources are MSA-only, and dialectal input degrades translation. The author first annotated a random Facebook sample and measured how a phrase-based MT system performs on each dialect. He then built several classifiers using word unigram features inside a log-linear framework:

- a weakly supervised model, with labels derived from the author's country in the user profile;
- a strongly supervised model, trained on manually labeled data;
- semi-supervised models, trained on large unlabeled pools that the first two classifiers labeled and filtered.

When all of them are combined, accuracy improves by roughly 5% absolute relative to the strongly supervised baseline and by 20% relative to the weakly supervised one. These gains hold on both the AOC corpus and a Facebook test set. On AOC, MSA vs. Egyptian accuracy reaches 92.0%, the best reported at that time. The combined classifier was then used to keep only MSA text for an English→Arabic language model. This shrank the model considerably and slightly improved BLEU, showing that dialect identification pays off downstream.

## Key points
- In 2,700 annotated Facebook sentences, MSA covered 58% and Egyptian 34%, followed by Levantine and Gulf, with Maghrebi rarest. An MSA-trained MT system scored about 18 BLEU on MSA but only about 10–12 on Egyptian and Levantine.
- Weak labels map the author's country to 5 dialect groups (Egyptian, Gulf, Levantine, Iraqi, Maghrebi). Unigram features outperformed bigram and trigram models.
- On AOC, MSA vs. EGY accuracy was 92.0%, compared with earlier results of 85.3% (Elfardy and Diab), 87.9% (Zaidan and Callison-Burch), and 89.1% (Tillmann et al.). No richer features were added.
- On the FB test set, the combined system helped MSA and Gulf but lost slightly on Egyptian and Levantine. The author attributes this to genre and topic mismatch between the training and test data.
- MSA filtering reduced the LM data from 200M to 30M sentences and memory from 23G to 6.6G. BLEU rose from 12.52 to 13.14 for English→Arabic.

## Interesting ideas
- Cheap metadata (the author's country) can serve as a noisy label source, giving a usable classifier without any hand annotation.
- Weak and strong classifiers can label a huge unlabeled pool, and only confident items are kept to retrain new models, so unlabeled data improves results without new features.
- Models trained on different supervision sources win on different classes, so combining them is more robust than picking one, especially under domain shift.
- Downstream quality was broken down per language variety (BLEU per dialect), which showed exactly where the system fails and where data effort pays off.
- A classifier was judged by its downstream effect, where filtering training text gave a smaller and better model rather than just higher accuracy.

## For Mova

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

- **I1** (learn): Source metadata (textbook vs. fable vs. forum origin) could bootstrap a register/genre tagger whose output then routes texts to suitable silver-data pipelines.
- **I2** (learn): This mirrors our witness-parser agreement; adding an explicit confidence threshold before a sentence enters silver training could reduce noise.
- **I3** (tree): Combining the perceptron parser and gparse per construction type, rather than globally, may help when evaluating on out-of-domain English.
- **I4** (tree): Reporting LAS separately for informal, dialectal, or converted-treebank English would show which UD conversions deserve investment.
- **I5** (back): Filtering generation and compressor training text to a single target register could shrink resources while making the generated output more consistent.

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

## Institutions

- [Meta](https://md.research.elamur.ai/free/en/companies/meta/) — Company, Menlo Park, United States · 183 works · 228,558 citations

📚 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: Fei Huang. 2015. Improved Arabic Dialect Classification with Social Media Data. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing.

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

```bibtex
@inproceedings{huang2015improved,
  title = {{Improved Arabic Dialect Classification with Social Media Data}},
  author = {Fei Huang},
  booktitle = {Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing},
  year = {2015},
  doi = {10.18653/v1/d15-1254},
  url = {https://doi.org/10.18653/v1/d15-1254}
}
```

- [← previous Automatic Dialect Detection in Arabic Broadcast Speech](https://md.research.elamur.ai/free/en/papers/ali-2015-automatic-dialect-detection-arabic-broadcast/)
- [next → Improving Arabic Diacritization through Syntactic Analysis](https://md.research.elamur.ai/free/en/papers/shahrour-2015-improving-arabic-diacritization-through-syntactic/)
- [↑ back to papers of 2015](https://md.research.elamur.ai/free/en/papers/2015/page/3/)

## References

19 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Avrim Blum and Tom Mitchell. 1998. [Combining labeled and unlabeled data with co-training](https://md.research.elamur.ai/free/en/papers/blum-1998-combining-labeled-unlabeled-data-co-training/). In Proceedings of the eleventh annual conference on Computational learning theory, pages 92–100. ACM. [↗](https://doi.org/10.1184/r1/6604187)
2. Ryan Cotterell and Chris Callison-Burch. 2014. [A multi-dialect, multi-genre corpus of informal written arabic](https://md.research.elamur.ai/free/en/papers/cotterell-2014-multi-dialect-multi-genre-corpus/). In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), Reykjavik, Iceland, may.
3. Kareem Darwish, Hassan Sajjad, and Hamdy Mubarak. 2014. [Verifiably effective arabic dialect identification](https://md.research.elamur.ai/free/en/papers/darwish-2014-verifiably-effective-arabic-dialect-identification/). In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1465–1468. Association for Computational Linguistics.
4. Mona Diab, Nizar Habash, Owen Rambow, Mohamed Altantawy, and Yassine Benajiba. 2010. Colaba: Arabic dialect annotation and processing. In LREC Workshop on Semitic Language Processing, pages 66–74. [find](https://scholar.google.com/scholar?q=Mona+Diab%2C+Nizar+Habash%2C+Owen+Rambow%2C+Mohamed+Altantawy%2C+and+Yassine+Benajiba%2E+2010%2E+Colaba%3A+Arabic+dialect+annotation+and+processing%2E+In+LREC+Workshop+on+Semitic+Language+Processing%2C+pages+66%E2%80%9374%2E)
5. Jacob Eisenstein. 2014. Identifying regional dialects in online social media. [find](https://scholar.google.com/scholar?q=Jacob+Eisenstein%2E+2014%2E+Identifying+regional+dialects+in+online+social+media%2E)
6. Heba Elfardy and Mona T Diab. 2012. [Simplified guidelines for the creation of large scale dialectal arabic annotations](https://md.research.elamur.ai/free/en/papers/elfardy-2012-simplified-guidelines-creation-large-scale/). In LREC, pages 371–378.
7. Heba Elfardy and Mona T Diab. 2013. [Sentence level dialect identification in arabic](https://md.research.elamur.ai/free/en/papers/elfardy-2013-sentence-level-dialect-identification-arabic/). In ACL (2), pages 456–461.
8. Nizar Habash and Owen Rambow. 2006. [Magead: a morphological analyzer and generator for the arabic dialects](https://md.research.elamur.ai/free/en/papers/habash-2006-magead-morphological-analyzer-generator-arabic/). In Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics, pages 681–688. Association for Computational Linguistics. [↗](https://doi.org/10.7916/d87s7x73)
9. Nizar Habash, Owen Rambow, Mona Diab, and Reem Kanjawi-Faraj. 2008. Guidelines for annotation of arabic dialectness. In Proceedings of the LREC Workshop on HLT & NLP within the Arabic world, pages 49–53. [find](https://scholar.google.com/scholar?q=Nizar+Habash%2C+Owen+Rambow%2C+Mona+Diab%2C+and+Reem+Kanjawi%2DFaraj%2E+2008%2E+Guidelines+for+annotation+of+arabic+dialectness%2E+In+Proceedings+of+the+LREC+Workshop+on+HLT+%26+NLP+within+the+Arabic+world%2C+pages+49%E2%80%93)
10. Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran, Richard Zens, Chris Dyer, Ondřej Bojar, Alexandra Constantin, and Evan Herbst. 2007. Moses: Open source toolkit for statistical machine translation. In Proceedings of the 45th Annual Meeting of the ACL on Interactive Poster and Demonstration Sessions, ACL ’07, pages 177–180, Stroudsburg, PA, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Philipp+Koehn%2C+Hieu+Hoang%2C+Alexandra+Birch%2C+Chris+Callison%2DBurch%2C+Marcello+Federico%2C+Nicola+Bertoldi%2C+Brooke+Cowan%2C+Wade+Shen%2C+Christine+Moran%2C+Richard+Zens%2C+Chris+Dyer%2C+Ondr%CC%8Cej+Bojar%2C+Alexandra+Const)
11. Yun Lei and John HL Hansen. 2011. Dialect classification via text-independent training and testing for arabic, spanish, and chinese. Audio, Speech, and Language Processing, IEEE Transactions on, 19(1):85– 96. [↗](https://doi.org/10.1109/tasl.2010.2045184)
12. Scott Novotney, Richard M Schwartz, and Sanjeev Khudanpur. 2011. Unsupervised arabic dialect adaptation with self-training. In INTERSPEECH, pages 541–544. Citeseer. [find](https://scholar.google.com/scholar?q=Scott+Novotney%2C+Richard+M+Schwartz%2C+and+Sanjeev+Khudanpur%2E+2011%2E+Unsupervised+arabic+dialect+adaptation+with+self%2Dtraining%2E+In+INTERSPEECH%2C+pages+541%E2%80%93544%2E+Citeseer%2E)
13. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. [Bleu: a method for automatic evaluation of machine translation](https://md.research.elamur.ai/free/en/papers/papineni-2002-bleu-method-automatic-evaluation-machine/). In Proceedings of the 40th annual meeting on association for computational linguistics, pages 311–318. Association for Computational Linguistics. [↗](https://doi.org/10.3115/1073083.1073135)
14. Wael Salloum and Nizar Habash. 2011. [Dialectal to standard arabic paraphrasing to improve arabic-english statistical machine translation](https://md.research.elamur.ai/free/en/papers/salloum-2011-dialectal-standard-arabic-paraphrasing-improve/). In Proceedings of the First Workshop on Algorithms and Resources for Modelling of Dialects and Language Varieties, pages 10–21. Association for Computational Linguistics. [↗](https://doi.org/10.7916/d8697bg2)
15. Hassan Sawaf. 2010. [Arabic dialect handling in hybrid machine translation](https://md.research.elamur.ai/free/en/papers/sawaf-2010-arabic-dialect-handling-hybrid-machine/). In Proceedings of the 9th Conference of the Association for Machine Translation in the Americas.
16. Christoph Tillmann, Saab Mansour, and Yaser Al-Onaizan, 2014. Proceedings of the First Workshop on Applying NLP Tools to Similar Languages, Varieties and Dialects, chapter [Improved Sentence-Level Arabic Dialect Classification](https://md.research.elamur.ai/free/en/papers/tillmann-2014-improved-sentence-level-arabic-dialect/), pages 110– 119. Association for Computational Linguistics and Dublin City University.
17. Omar F. Zaidan and Chris Callison-Burch. 2011. [The arabic online commentary dataset: an annotated dataset of informal arabic with high dialectal content](https://md.research.elamur.ai/free/en/papers/zaidan-2011-arabic-online-commentary-dataset-annotated/). In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 37–41, Portland, Oregon, USA, June. Association for Computational Linguistics.
18. 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)
19. Rabih Zbib, Erika Malchiodi, Jacob Devlin, David Stallard, Spyros Matsoukas, Richard Schwartz, John Makhoul, Omar F Zaidan, and Chris Callison-Burch. 2012. [Machine translation of arabic dialects](https://md.research.elamur.ai/free/en/papers/zbib-2012-machine-translation-arabic-dialects/). In Proceedings of the 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 49–59. Association for Computational Linguistics.

*Updated 1 November 2026*