Elamur Research⌕
☰
home / papers / 2016 / Creating Resources for Dialectal Arabic from a Single Annotation: A Case Study on Egyptian and Levantine

Creating Resources for Dialectal Arabic from a Single Annotation: A Case Study on Egyptian and Levantine

Paper · 2016 · Ramy Eskander, Nizar Habash, Owen Rambow et al.

Summary

Arabic dialects lack standard spelling, linguistic documentation and annotated data, yet they are increasingly written on social media. The authors propose a recipe for building both a morphological analyzer (all possible readings of a word, out of context) and a morphological tagger (the right reading in context) for a dialect. The recipe needs only one kind of annotation: running text where each word gets a CODA-normalized spelling, a segmentation, features and a lemma. From this annotation they build inflectional classes per lemma. They then automatically fill in the unseen paradigm slots ("paradigm completion"). Optionally, they merge in analyzers for related variants: MSA, and Egyptian for Levantine. A classifier-based analyze-and-choose tagger picks among the candidate analyses. They test this on Egyptian, using subsets that simulate low resources, and on Levantine, using the partially annotated Curras corpus. With modest data the system competes with or beats MADAMIRA-EGY. The work yields what the authors describe as the first Levantine analyzer and tagger. The method is meant to carry over to other related low-resource dialects.

Key points

  • Egyptian data: LDC Egyptian corpora with the Diab et al. (2013) splits.
    • DEV has 45K words and TEST has 46K words.
    • Only up to 135K words of TRAIN are used, with learning curves at 5K, 15K, 45K and 135K.
    • Annotations were aligned to the closest CALIMA Egy analysis. Words with no analysis fell back to proper noun, which happened for 4.9% of words.
  • Levantine data: the Curras Corpus of Palestinian Arabic.
    • It has about 57K words, half from a TV show transcript and half from social media and web text.
    • The version used was only 82% manually annotated. Split: TRAIN 45K, DEV 6K, TEST 6K.
    • DEV and TEST are fully annotated.
  • Egyptian results: the system matches or beats MADAMIRA-EGY on several metrics.
    • It wins on Lemma and All with just 15K training words, and on Stem with 45K.
    • With 135K words it also wins on POS5 and CODA, but never on full POS.
    • Gains from 45K to 135K are much smaller than the earlier increments.
  • Levantine results: ParaFill Lev +CALIMA Egy +SAMA ext beats the MADAMIRA-EGY baseline already at 5K training words (except on POS).
    • It consistently outperforms ParaFill Lev alone.
    • This holds even though ambiguity jumps from 2.2 to 14.2 analyses per word at 15K training words.
  • Error analysis (45K training, DEV):
    • For Egyptian, 20.7% of errors come from the gold analysis missing in the analyzer, 5.6% from back-off cases and 2.6% from OOV words.
    • For Levantine, 21.7% come from a missing gold analysis and 6.6% from OOV words.
    • All remaining errors are the tagger choosing wrongly among available analyses.

Interesting ideas

  • I1. Build an analyzer from a few observed forms: take the forms observed for each lemma and arrange them into a slot table per inflectional class, then complete the empty slots by analogy with other lemmas of the same class. This turns sparse annotation into a full paradigm lexicon.
  • I2. Represent stems as templates: replace most stem letters with placeholders but keep the "weak" letters that change during inflection. This is a cheap way to model templatic morphology and generalize patterns across lemmas.
  • I3. Allow partial annotation: the corpus does not need every word annotated, so annotators can focus on the most frequent words and still feed the pipeline.
  • I4. Over-generate candidates, then disambiguate: borrow candidates generously from related variants, even at the cost of much higher ambiguity, and let a context-aware chooser filter them. This beat a smaller, cleaner candidate set.
  • I5. Split errors by where they come from: separate the errors where the right answer was never offered (analyzer gap, back-off, OOV) from those where it was offered but not chosen (tagger fault). This shows which component to fix.

For Mova

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

  • I1 (tokens): Unseen English inflections (irregular plurals, -ed/-ing spelling changes) could be predicted by filling paradigm slots from lemmas that share a class.
  • I2 (tokens): Abstract stem shapes (keeping only the letters that alternate, like y/i or doubled consonants) could let the lemmatizer learn spelling-change rules that generalize instead of memorizing words.
  • I3 (learn): Silver-data training could accept partially labeled sentences, spending LLM-teacher or human effort only on frequent or contested tokens.
  • I4 (tree): When converting UD dialects or treebanks, candidate analyses from several sources could be pooled and a trained chooser selects one, rather than trusting a single narrow source.
  • I5 (learn): Evaluation could report, for each stage, whether the gold option existed among the candidates, so repair rules target the generator or the selector correctly.

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 UniversityNew York · United States241 works · 54,053 citations🎓New York University Abu DhabiAbu Dhabi · United Arab Emirates122 works · 4,602 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 2,908 worksMorphology · 405 works

Cite

Cite as Ramy Eskander, Nizar Habash, Owen Rambow and Arfath Pasha. 2016. Creating Resources for Dialectal Arabic from a Single Annotation: A Case Study on Egyptian and Levantine. Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers.

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

@inproceedings{eskander2016creating,
  title = {{Creating Resources for Dialectal Arabic from a Single Annotation: A Case Study on Egyptian and Levantine}},
  author = {Ramy Eskander, Nizar Habash, Owen Rambow and Arfath Pasha},
  booktitle = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers},
  year = {2016},
  url = {https://aclanthology.org/C16-1326/}
}

References

38 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Hitham Abo Bakr, Khaled Shaalan, and Ibrahim Ziedan. 2008. A Hybrid Approach for Converting Written Egyptian Colloquial Dialect into Diacritized Arabic. In The 6th International Conference on Informatics and Systems, INFOS2008. Cairo University. find
  2. Rania Al-Sabbagh and Roxana Girju. 2012a. A supervised POS Tagger for Written Arabic Social Networking Corpora. In Jeremy Jancsary, editor, Proceedings of KONVENS 2012, pages 39–52. ÖGAI, September. Main track: oral presentations. find
  3. Rania Al-Sabbagh and Roxana Girju. 2012b. YADAC: Yet another Dialectal Arabic Corpus. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC-2012), pages 2882–2889.
  4. Faisal Al-Shargi, Aidan Kaplan, Ramy Eskander, Nizar Habash, and Owen Rambow. 2016. Morphologically Annotated Corpora and Morphological Analyzers for Moroccan and Sanaani Yemeni Arabic. In 10th Language Resources and Evaluation Conference (LREC 2016).
  5. Houda Bouamor, Nizar Habash, and Kemal Oflazer. 2014. A Multidialectal Parallel Corpus of Arabic. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC-2014). European Language Resources Association (ELRA).
  6. Eric Brill. 1995. Transformation-Based Error-Driven Learning and Natural Language Processing: A Case Study in Part-of-Speech Tagging. Computational Linguistics, 21(4):543–565.
  7. Tim Buckwalter. 2004. Buckwalter Arabic Morphological Analyzer Version 2.0. LDC2004L02, ISBN 1-58563-324-0. LDC catalog number David Chiang, Mona Diab, Nizar Habash, Owen Rambow, and Safiullah Shareef. 2006. Parsing Arabic Dialects. In Proceedings of EACL, Trento, Italy. ↗
  8. Walter Daelemans and Antal van den Bosch. 2005. Memory-Based Language Processing. Studies in Natural Language Processing. Cambridge University Press, Cambridge, UK. ↗
  9. Mona Diab, Kadri Hacioglu, and Daniel Jurafsky. 2007. Automated Methods for Processing Arabic Text: From Tokenization to Base Phrase Chunking. In Antal van den Bosch and Abdelhadi Soudi, editors, Arabic Computational Morphology: Knowledge-based and Empirical Methods. Kluwer/Springer. find
  10. 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
  11. Mona Diab, Nizar Habash, Owen Rambow, and Ryan Roth. 2013. LDC Arabic Treebanks and Associated Corpora: Data Divisions Manual. arXiv preprint arXiv:1309.5652. ↗
  12. Kevin Duh and Katrin Kirchhoff. 2005. POS Tagging of Dialectal Arabic: a Minimally Supervised Approach. In Proceedings of the ACL Workshop on Computational Approaches to Semitic Languages, Semitic ’05, pages 55–62, Ann Arbor, Michigan.
  13. Ramy Eskander, Nizar Habash, Ann Bies, Seth Kulick, and Mohamed Maamouri. 2013a. Automatic Correction and Extension of Morphological Annotations. In Proceedings of the 7th Linguistic Annotation Workshop & Interoperability with Discourse, pages 1–10. find
  14. Ramy Eskander, Nizar Habash, and Owen Rambow. 2013b. Automatic Extraction of Morphological Lexicons from Morphologically Annotated Corpora. In Proceedings of tenth Conference on Empirical Methods in Natural Language Processing. find
  15. Hassan Gadalla, Hanaa Kilany, Howaida Arram, Ashraf Yacoub, Alaa El-Habashi, Amr Shalaby, Krisjanis Karins, Everett Rowson, Robert MacIntyre, Paul Kingsbury, David Graff, and Cynthia McLemore. 1997. CALLHOME Egyptian Arabic Transcripts. Linguistic Data Consortium, Philadelphia. find
  16. David Graff, Mohamed Maamouri, Basma Bouziri, Sondos Krouna, Seth Kulick, and Tim Buckwalter. 2009. Standard Arabic Morphological Analyzer (SAMA) Version 3.1. Linguistic Data Consortium LDC2009E73. find
  17. Nizar Habash and Owen Rambow. 2005. Arabic Tokenization, Part-of-Speech Tagging and Morphological Disambiguation in One Fell Swoop. In Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL’05), pages 573–580, Ann Arbor, Michigan.
  18. Nizar Habash, Abdelhadi Soudi, and Tim Buckwalter. 2007. On Arabic Transliteration. In A. van den Bosch and A. Soudi, editors, Arabic Computational Morphology: Knowledge-based and Empirical Methods. Springer. ↗
  19. Nizar Habash, Reem Faraj, and Ryan Roth. 2009. Syntactic Annotation in the Columbia Arabic Treebank. In Proceedings of MEDAR International Conference on Arabic Language Resources and Tools, Cairo, Egypt. find
  20. Nizar Habash, Mona Diab, and Owen Rabmow. 2012a. Conventional Orthography for Dialectal Arabic. In Proceedings of LREC, Istanbul, Turkey.
  21. Nizar Habash, Ramy Eskander, and Abdelati Hawwari. 2012b. A Morphological Analyzer for Egyptian Arabic. In NAACL-HLT 2012 Workshop on Computational Morphology and Phonology (SIGMORPHON2012), pages 1–9.
  22. Nizar Habash, Ryan Roth, Owen Rambow, Ramy Eskander, and Nadi Tomeh. 2013. Morphological Analysis and Disambiguation for Dialectal Arabic. In Proceedings of the 2013 Conference of NAACL-HLT, Atlanta, GA.
  23. Nizar Habash. 2007. Arabic Morphological Representations for Machine Translation. In Antal van den Bosch and Abdelhadi Soudi, editors, Arabic Computational Morphology: Knowledge-based and Empirical Methods. Kluwer/Springer. ↗
  24. Nizar Habash. 2010. Introduction to Arabic Natural Language Processing. Morgan & Claypool Publishers. ↗
  25. Jan Hajič. 2000. Morphological Tagging: Data vs. Dictionaries. In Proceedings of the 1st Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL’00), Seattle, WA.
  26. Mustafa Jarrar, Nizar Habash, Diyam Akra, and Nasser Zalmout. 2014. Building a Corpus for Palestinian Arabic: a Preliminary Study. In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP), pages 18–27, Doha, Qatar, October. Association for Computational Linguistics.
  27. Salam Khalifa, Nizar Habash, Dana Abdulrahim, and Sara Hassan. 2016. A Large Scale Corpus of Gulf Arabic. In Proceedings of the International Conference on Language Resources and Evaluation (LREC), Portorož, Slovenia.
  28. H. Kilany, H. Gadalla, H. Arram, A. Yacoub, A. El-Habashi, and C. McLemore. 2002. Egyptian Colloquial Arabic Lexicon. LDC catalog number LDC99L22. find
  29. Mohamed Maamouri, Ann Bies, Seth Kulick, Dalila Tabessi, and Sondos Krouna. 2012. Egyptian Arabic Treebank DF Parts 1-8 V2.0 - LDC catalog numbers LDC2012E93, LDC2012E98, LDC2012E89, LDC2012E99, LDC2012E107, LDC2012E125, LDC2013E12, LDC2013E21. find
  30. Mohamed Maamouri, Ann Bies, Seth Kulick, Michael Ciul, Nizar Habash, and Ramy Eskander. 2014. Developing an Egyptian Arabic Treebank: Impact of Dialectal Morphology on Annotation and Tool Development. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC-2014). European Language Resources Association (ELRA). find
  31. Abir Masmoudi, Mariem Ellouze Khmekhem, Yannick Esteve, Lamia Hadrich Belguith, and Nizar Habash. 2014. A Corpus and Phonetic Dictionary for Tunisian Arabic Speech Recognition. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC-2014). European Language Resources Association (ELRA).
  32. Emad Mohamed, Behrang Mohit, and Kemal Oflazer. 2012. Annotating and Learning Morphological Segmentation of Egyptian Colloquial Arabic. In Proceedings of the Language Resources and Evaluation Conference (LREC), Istanbul.
  33. Arfath Pasha, Mohamed Al-Badrashiny, Ahmed El Kholy, Ramy Eskander, Mona Diab, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. MADAMIRA: A Fast, Comprehensive Tool for Morphological Analysis and Disambiguation of Arabic. In In Proceedings of LREC, Reykjavik, Iceland.
  34. 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, Edinburgh, Scotland. ↗
  35. Wael Salloum and Nizar Habash. 2014. ADAM: Analyzer for Dialectal Arabic Morphology. Journal of King Saud University-Computer and Information Sciences, 26(4):372–378. ↗
  36. Kamel Smaı̈li, Mourad Abbas, Karima Meftouh, and Salima Harrat. 2014. Building Resources for Algerian Arabic Dialects. In 15th Annual Conference of the International Communication Association Interspeech. find
  37. Clare Voss, Stephen Tratz, Jamal Laoudi, and Douglas Briesch. 2014. Finding Romanized Arabic Dialect in Code-Mixed Tweets. In Nicoletta Calzolari, Khalid Choukri, Thierry Declerck, Hrafn Loftsson, Bente Maegaard, Joseph Mariani, Asuncion Moreno, Jan Odijk, and Stelios Piperidis, editors, Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 2249–2253, Reykjavik, Iceland, May. European Language Resources Association (ELRA). ACL Anthology Identifier: L14-1086. ↗
  38. Imed Zitouni, Jeffrey S. Sorensen, and Ruhi Sarikaya. 2006. Maximum Entropy Based Restoration of Arabic Diacritics. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, pages 577–584, Sydney, Australia.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/eskander-2016-creating-resources-dialectal-arabic-single/