---
title: "Sentence Level Dialect Identification in Arabic — Elfardy 2013"
description: "The paper tackles sentence-level dialect identification for Arabic. 85.5% accuracy. 14 references, cited by 74 papers here."
authors: ["Heba Elfardy", "Mona Diab"]
year: 2013
venue: "Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)"
source: "https://aclanthology.org/P13-2081/"
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/elfardy-2013-sentence-level-dialect-identification-arabic/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/elfardy-2013-sentence-level-dialect-identification-arabic/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2013](https://md.research.elamur.ai/free/en/papers/2013/) / Sentence Level Dialect Identification in Arabic

# Sentence Level Dialect Identification in Arabic

**Authors:** Heba Elfardy and [Mona Diab](https://md.research.elamur.ai/free/en/authors/mona-diab/) · **Year:** 2013 · **Venue:** [Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2013](https://md.research.elamur.ai/free/en/venues/acl/2013/)) · **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/P13-2081/) · [PDF](https://aclanthology.org/P13-2081.pdf) · [OpenAlex](https://openalex.org/W2251986002)

**Citations:** 117 citations (OpenAlex, 2026-10-14) · [74](https://md.research.elamur.ai/free/en/papers/elfardy-2013-sentence-level-dialect-identification-arabic/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
The paper tackles sentence-level dialect identification for Arabic. Specifically, it decides whether a sentence is mainly Modern Standard Arabic (MSA) or Egyptian Dialectal Arabic (EDA). This matters because tools built for MSA break down on dialectal or code-switched text, so a reliable classifier is needed as a preprocessing step. The authors start from an existing token-level dialect tagger (Elfardy et al., 2013) and aggregate its per-word labels into sentence features. They add three more kinds of evidence:
- perplexities from MSA and EDA language models,
- morphological-analyzer coverage,
- informality "meta" cues.

All features feed a Naive Bayes classifier. They test four text preprocessing variants for the LMs: surface, tokenized, CODA-normalized, and their combination. Experiments use the code-switched commentary data of Zaidan and Callison-Burch (2011). The tokenized setup reaches 85.5% accuracy under 10-fold cross-validation, beating the earlier LM-based approach (80.9%) and the authors' own token-threshold and perplexity baselines.

## Key points
- **Data:** the EDA-MSA portion of the Zaidan and Callison-Burch (2011) crowd-annotated corpus of comments on Egyptian news.
  - Training set: about 12.2K MSA and 11.3K EDA sentences.
  - Held-out test set: about 1.35K MSA and 1.25K EDA sentences.
- **Classifier and features:** Naive Bayes in WEKA. Core features are the share of EDA and OOV tokens, MSA/EDA analyzer coverage, EDA-lexicon coverage, and MSA vs. EDA LM perplexity. Meta features include punctuation, emoticons, exclamation marks, word lengthening, and Roman-script words.
- **Preprocessing:** tokenization uses the D3 scheme via MADA (for MSA) and MADA-ARZ (for EDA), with 5-gram LMs. Orthographic normalization uses CODAfy. Surface LMs are 3-gram.
- **LM size:** learning curves cover LM sizes from 2M to 28M tokens, and 8M worked best.
- **Results in the setting comparable to prior work** (accuracy, 10-fold CV on the whole dataset):

  | System | Accuracy |
  |---|---|
  | Majority baseline | 51.9% |
  | Token-threshold baseline | 78.5% |
  | Perplexity baseline | 80.4% |
  | Zaidan and Callison-Burch (2011) | 80.9% |
  | Surface | 82.6% |
  | CODA | 82.8% |
  | Tokenized | 85.5% |
  | Tokenized-CODA | 84.9% |

  On the held-out test set, the Tokenized configuration scores 83.3%.

## Interesting ideas
- **Lift word decisions to sentence features.** Turning per-token language labels into sentence-level proportions (share dialectal, share OOV) cheaply reuses a word-level tagger for a coarser task. The same trick could serve other code-switching settings, such as Ukrainian–Russian surzhyk detection.
- **Normalization can erase the signal you need.** Orthographic normalization (CODA) reduces sparsity but also maps dialect-specific spellings back to standard forms. Tokenization fixed sparsity without losing those cues, and stacking both actually hurt. It is worth checking whether preprocessing removes the very differences a classifier relies on.
- **Cap repeated letters instead of deleting them.** The authors reduce elongated spellings to exactly three repeated letters. This keeps a normalized form while preserving the hint that an informal speech effect was present. It is a useful middle ground for any noisy social-media text.
- **Use informality features as dialect proxies.** Emoticons, repeated punctuation, and Latin-script words say nothing about specific words, but they signal informal register, which correlates with dialect use. This kind of style evidence is easy to add to other variety- or register-identification tasks.
- **Report two evaluation protocols.** The authors give results both on a held-out test split and under cross-validation on the full dataset. The first better reflects generalization to unseen data, while only the second is directly comparable to prior work. Being explicit about which numbers are comparable is good practice.

*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

- [Columbia University](https://md.research.elamur.ai/free/en/universities/columbia-university/) — University, New York, United States · 149 works · 48,852 citations
- [George Washington University](https://md.research.elamur.ai/free/en/universities/george-washington-university/) — University, Washington, United States · 21 works · 7,106 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

🧭 Topics (1): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 1,484 works

## Cite

Cite as: Heba Elfardy and Mona Diab. 2013. Sentence Level Dialect Identification in Arabic. Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers).

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

```bibtex
@inproceedings{elfardy2013sentence,
  title = {{Sentence Level Dialect Identification in Arabic}},
  author = {Heba Elfardy and Mona Diab},
  booktitle = {Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)},
  year = {2013},
  url = {https://aclanthology.org/P13-2081/}
}
```

- [← previous A Dataset of Syntactic-Ngrams over Time from a Very Large Corpus of English Books](https://md.research.elamur.ai/free/en/papers/goldberg-2013-dataset-syntactic-ngrams-over-time/)
- [next → Target Language Adaptation of Discriminative Transfer Parsers](https://md.research.elamur.ai/free/en/papers/tackstrom-2013-target-language-adaptation-discriminative-transfer/)
- [↑ back to papers of 2013](https://md.research.elamur.ai/free/en/papers/2013/)

## References

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

1. Fadi Biadsy, Julia Hirschberg, and Nizar Habash. 2009. Spoken arabic dialect identification using phonotactic modeling. In Proceedings of the Workshop on Computational Approaches to Semitic Languages at the meeting of the European Association for Computational Linguistics (EACL), Athens, Greece. [find](https://scholar.google.com/scholar?q=Fadi+Biadsy%2C+Julia+Hirschberg%2C+and+Nizar+Habash%2E+2009%2E+Spoken+arabic+dialect+identification+using+phonotactic+modeling%2E+In+Proceedings+of+the+Workshop+on+Computational+Approaches+to+Semitic+Languages+)
2. Pradeep Dasigi and Mona Diab. 2011. Codact: Towards identifying orthographic variants in dialectal arabic. In Proceedings of the 5th International Joint Conference on Natural Language Processing (ICJNLP), Chiangmai, Thailand. [↗](https://doi.org/10.7916/d8np2dqz)
3. Heba Elfardy and Mona Diab. 2012a. Simplified guidelines for the creation of large scale dialectal arabic annotations. In Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC), Istanbul, Turkey. [find](https://scholar.google.com/scholar?q=Heba+Elfardy+and+Mona+Diab%2E+2012a%2E+Simplified+guidelines+for+the+creation+of+large+scale+dialectal+arabic+annotations%2E+In+Proceedings+of+the+8th+International+Conference+on+Language+Resources+and+Eval)
4. Heba Elfardy and Mona Diab. 2012b. Token level identification of linguistic code switching. In Proceedings of the 24th International Conference on Computational Linguistics (COLING),Mumbai, India. [find](https://scholar.google.com/scholar?q=Heba+Elfardy+and+Mona+Diab%2E+2012b%2E+Token+level+identification+of+linguistic+code+switching%2E+In+Proceedings+of+the+24th+International+Conference+on+Computational+Linguistics+%28COLING%29%2CMumbai%2C+India%2E)
5. Heba Elfardy, Mohamed Al-Badrashiny, and Mona Diab. 2013. Code Switch Point Detection in Arabic. In Proceedings of the 18th International Conference on Application of Natural Language to Information Systems (NLDB2013), MediaCity, UK, June. [find](https://scholar.google.com/scholar?q=Heba+Elfardy%2C+Mohamed+Al%2DBadrashiny%2C+and+Mona+Diab%2E+2013%2E+Code+Switch+Point+Detection+in+Arabic%2E+In+Proceedings+of+the+18th+International+Conference+on+Application+of+Natural+Language+to+Information+S)
6. Ramy Eskander, Nizar Habash, Owen Rambow, and Nadi Tomeh. 2013. Processing Spontaneous Orthography. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), Atlanta, GA. [find](https://scholar.google.com/scholar?q=Ramy+Eskander%2C+Nizar+Habash%2C+Owen+Rambow%2C+and+Nadi+Tomeh%2E+2013%2E+Processing+Spontaneous+Orthography%2E+In+Proceedings+of+the+2013+Conference+of+the+North+American+Chapter+of+the+Association+for+Computati)
7. Ferguson. 1959. Diglossia. Word 15. 325340. [find](https://scholar.google.com/scholar?q=Ferguson%2E+1959%2E+Diglossia%2E+Word+15%2E+325340%2E)
8. Nizar Habash, Owen Rambow, and Ryan Roth. 2009. Mada+ tokan: A toolkit for arabic tokenization, diacritization, morphological disambiguation, pos tagging, stemming and lemmatization. In Proceedings of the 2nd International Conference on Arabic Language Resources and Tools (MEDAR), Cairo, Egypt, pages 102–109. [find](https://scholar.google.com/scholar?q=Nizar+Habash%2C+Owen+Rambow%2C+and+Ryan+Roth%2E+2009%2E+Mada%2B+tokan%3A+A+toolkit+for+arabic+tokenization%2C+diacritization%2C+morphological+disambiguation%2C+pos+tagging%2C+stemming+and+lemmatization%2E+In+Proceedings+of)
9. Nizar Habash, Mona Diab, and Owen Rabmow. 2012. [Conventional orthography for dialectal arabic](https://md.research.elamur.ai/free/en/papers/habash-2012-conventional-orthography-dialectal-arabic/). In Proceedings of the Language Resources and Evaluation Conference (LREC), Istanbul.
10. Nizar Habash, Ryan Roth, Owen Rambow, Ramy Eskander, and Nadi Tomeh. 2013. [Morphological Analysis and Disambiguation for Dialectal Arabic](https://md.research.elamur.ai/free/en/papers/habash-2013-morphological-analysis-disambiguation-dialectal-arabic/). In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), Atlanta, GA.
11. Nizar Habash. 2010. Introduction to arabic natural language processing. Advances in neural information processing systems. [↗](https://doi.org/10.1007/978-3-031-02139-8)
12. Mark Hall, Eibe Frank, Geoffrey Holmes, Bernhard Pfahringer, Peter Reutemann, and Ian H Witten. 2009. The weka data mining software: an update. ACM SIGKDD Explorations Newsletter, 11(1):10– 18. [↗](https://doi.org/10.1145/1656274.1656278)
13. Wael Salloum and Nizar Habash. 2011. Dialectal to standard arabic paraphrasing to improve arabic-english statistical machine translation. 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)
14. 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 ACL, pages 37–41.

*Updated 24 October 2026*