---
title: "Transforming Standard Arabic to Colloquial Arabic (2012)"
description: "The authors propose to generate CEA text automatically from MSA text whose morphology has… 7 references, cited by 13 papers here."
authors: ["Emad Mohamed", "Behrang Mohit", "Kemal Oflazer"]
year: 2012
venue: "Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)"
source: "https://aclanthology.org/P12-2035/"
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/mohamed-2012-transforming-standard-arabic-colloquial-arabic/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/mohamed-2012-transforming-standard-arabic-colloquial-arabic/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2012](https://md.research.elamur.ai/free/en/papers/2012/page/2/) / Transforming Standard Arabic to Colloquial Arabic

# Transforming Standard Arabic to Colloquial Arabic

**Authors:** [Emad Mohamed](https://md.research.elamur.ai/free/en/authors/emad-mohamed/), [Behrang Mohit](https://md.research.elamur.ai/free/en/authors/behrang-mohit/) and [Kemal Oflazer](https://md.research.elamur.ai/free/en/authors/kemal-oflazer/) · **Year:** 2012 · **Venue:** [Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2012](https://md.research.elamur.ai/free/en/venues/acl/2012/)) · **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/P12-2035/) · [PDF](https://aclanthology.org/P12-2035.pdf)

**Citations:** [13](https://md.research.elamur.ai/free/en/papers/mohamed-2012-transforming-standard-arabic-colloquial-arabic/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Arabic NLP tools are mostly built for Modern Standard Arabic (MSA), while dialects such as Colloquial Egyptian Arabic (CEA) have few annotated resources. The authors propose to generate CEA text automatically from MSA text whose morphology has already been disambiguated. Their key observation is that the two varieties differ mostly in affixes and function words, while stems usually stay the same. They wrote 103 hand-crafted rules over segment-level POS tags, covering nominal, verbal, pronominal and demonstrative morphology, plus about 400 lexical substitutions. Because CEA spelling is not standardized, one MSA form can expand into many valid colloquial spellings. They applied the rules to two Arabic Treebank sections and trained a memory-based POS tagger on the converted data. On real Egyptian web comments, tagging accuracy rose from 73.24% to 86.84%, and the unknown-word rate fell from 28.98% to 16.66%. This shows that existing standard-language annotations can be reused cheaply to bootstrap tools for a closely related dialect.

## Key points
- The rule set has 103 morphological rules (number/case affixes, tense, modality, gender, dual-to-plural collapse, clitic negation m…$) plus roughly 400 MSA→CEA lexical mappings. A single negated imperfect verb can yield 24 legitimate CEA spellings.
- Training data: ATB sections p2v3 and p3v2, converted to CEA and tagged with whole-word composite tags from the Columbia Arabic Treebank 6-tag set (58 composite tags). The tagger is MBT (IGTree for known words, IB1 for unknown words, MVDM, k=25).
- Evaluation data: hand-annotated, highly colloquial masrawy.com comments, with a dev set of 100 comments (2,757 words) and a test set of 192 comments (7,092 words).
- Choosing one random valid variant per sentence (79.67% test) beat choosing the variant ranked best by a trigram SRILM model trained on 11.5M words of AOC comments (74.32%). The random setting also produced a larger vocabulary: 55,367 versus 51,306 word types.
- Test accuracy rose step by step: MSA-only 73.24%, then hybrid MSA plus converted data 83.81%, then hybrid plus dev data 86.84%.

## Interesting ideas
- Instead of adapting the model, the authors rewrite already-annotated training text into the target variety, so the gold labels carry over for free.
- When several outputs are all valid, picking one at random beat picking the most probable one, because random choice spreads coverage across variants.
- Mixing original-variety data with converted data, then adding a small amount of authentic in-domain gold, gave gains at each step; the final step exposed what the conversion had missed.
- The conversion relies on segment-level disambiguation first, because surface affixes are ambiguous; only function morphemes are rewritten, while stems stay put.
- Accuracy is reported separately for known and unknown words, together with the unknown-word rate, which shows whether gains come from coverage or from better context modelling.

## For Mova

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

- **I1** (tree): Rule-based rewriting of an English UD treebank into informal or social-media English (contractions, dropped subjects, nonstandard spellings) could give the parsers labeled noisy-text training data while keeping the trees intact.
- **I2** (learn): When generating silver or augmented data, sample among acceptable variants rather than always taking the top-scored one, so the training set covers more surface forms.
- **I3** (learn): Train on original plus transformed corpora, keep a small real-domain set, and use the gap it closes as a diagnostic list of phenomena the transformation rules still miss.
- **I4** (back): A generator that renders text in different registers could likewise rewrite only function words and inflections over an already-analyzed graph, leaving content lemmas fixed for explainability.
- **I5** (learn): Splitting tagger and parser evaluation into known/unknown-token slices with OOV rates would show whether data changes help coverage or disambiguation.

*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 (2): [🖥️ 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

🧭 Topics (2): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works · [🧩 Morphology](https://md.research.elamur.ai/free/en/topics/morphology/) · 405 works

## Cite

Cite as: Emad Mohamed, Behrang Mohit and Kemal Oflazer. 2012. Transforming Standard Arabic to Colloquial Arabic. Proceedings of the 50th 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{mohamed2012transforming,
  title = {{Transforming Standard Arabic to Colloquial Arabic}},
  author = {Emad Mohamed, Behrang Mohit and Kemal Oflazer},
  booktitle = {Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)},
  year = {2012},
  url = {https://aclanthology.org/P12-2035/}
}
```

- [← previous The Parallel-TUT: a multilingual and multiformat treebank](https://md.research.elamur.ai/free/en/papers/bosco-2012-parallel-tut-multilingual-multiformat-treebank/)
- [next → Translating English Discourse Connectives into Arabic: a Corpus-based Analysis and an Evaluation Metric](https://md.research.elamur.ai/free/en/papers/hajlaoui-2012-translating-english-discourse-connectives-into/)
- [↑ back to papers of 2012](https://md.research.elamur.ai/free/en/papers/2012/page/2/)

## References

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

1. Bies, Ann and Maamouri, Mohamed (2003). Penn Arabic Treebank guidelines. Technical report, LDC, University of Pennsylvania. [find](https://scholar.google.com/scholar?q=Bies%2C+Ann+and+Maamouri%2C+Mohamed+%282003%29%2E+Penn+Arabic+Treebank+guidelines%2E+Technical+report%2C+LDC%2C+University+of+Pennsylvania%2E)
2. Buckwalter, T. (2002). Arabic Morphological Analyzer (AraMorph). Version 1.0. Linguistic Data Consortium, catalog number LDC2002L49 and ISBN 1-58563-257- 0 Daelemans, Walter and van den Bosch, Antal ( 2005). [find](https://scholar.google.com/scholar?q=Buckwalter%2C+T%2E+%282002%29%2E+Arabic+Morphological+Analyzer+%28AraMorph%29%2E+Version+1%2E0%2E+Linguistic+Data+Consortium%2C+catalog+number+LDC2002L49+and+ISBN+1%2D58563%2D257%2D+0+Daelemans%2C+Walter+and+van+den+Bosch%2C+Antal+%28)
3. Memory Based Language Processing. Cambridge University Press. Daelemans, Walter; Zavrel, Jakub; Berck, Peter, and Steven Gillis (1996). [MBT: A memory-based part of speech tagger-generator](https://md.research.elamur.ai/free/en/papers/daelemans-1996-mbt-memory-based-part-speech/). In Eva Ejerhed and Ido Dagan, editors, Proceedings of the 4th Workshop on Very Large Corpora, pages 14–27, Copenhagen, Denmark. Diab, Mona; Habash, Nizar; Rambow, Owen; Altantawy, Mohamed, and Benajiba, Yassine. COLABA: Arabic Dialect Annotation and Processing. LREC 2010.
4. Duh, K. and Kirchhoff, K. (2005). [POS Tagging of Dialectal Arabic: A Minimally Supervised Approach](https://md.research.elamur.ai/free/en/papers/duh-2005-pos-tagging-dialectal-arabic-minimally/).
5. Proceedings of the ACL Workshop on Computational Approaches to Semitic Languages, Ann Arbor, June 2005. [find](https://scholar.google.com/scholar?q=Proceedings+of+the+ACL+Workshop+on+Computational+Approaches+to+Semitic+Languages%2C+Ann+Arbor%2C+June+2005%2E)
6. Habash, Nizar; Rambow, Own and Kiraz, George (2005). [Morphological analysis and generation for Arabic dialects](https://md.research.elamur.ai/free/en/papers/habash-2005-morphological-analysis-generation-arabic-dialects/). Proceedings of the ACL Workshop on Computational Approaches to Semitic Languages, pages 17–24, Ann Arbor, June 2005 Habash, Nizar and Roth, Ryan. CATiB: The Columbia Arabic Treebank. Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 221–224, Singapore, 4 August 2009. c 2009 ACL and AFNLP Habash, Nizar, Owen Rambow and Ryan Roth. MA-DA+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, 2009 Kundu, Gourab abd Roth, Don (2011). Adapting Text instead of the Model: An Open Domain Approach. Proceedings of the Fifteenth Conference on Computational Natural Language Learning, pages 229–237,Portland, Oregon, USA, 23–24 June 2011 Mohamed, Emad. and Kuebler, Sandra (2010). Is Arabic Part of Speech Tagging Feasible Without Word Segmentation? Proceedings of HLT-NAACL 2010, Los Angeles, CA.
7. Stolcke, A. (2002). SRILM - an extensible language modeling toolkit. In Proc. of ICSLP, Denver, Colorado [find](https://scholar.google.com/scholar?q=Stolcke%2C+A%2E+%282002%29%2E+SRILM+%2D+an+extensible+language+modeling+toolkit%2E+In+Proc%2E+of+ICSLP%2C+Denver%2C+Colorado)

*Updated 30 October 2026*