---
title: "Repérage des entités nommées pour l'arabe : adaptation"
description: "The paper builds a statistical named entity recognizer for Arabic. 26 references."
authors: ["Souhir Gahbiche-Braham", "Hélène Bonneau-Maynard", "Thomas Lavergne", "François Yvon"]
year: 2012
venue: "Proceedings of the Joint Conference JEP-TALN-RECITAL 2012, volume 2: TALN"
source: "https://aclanthology.org/F12-2044/"
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/braham-2012-reperage-des-entites-nommees-pour/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/braham-2012-reperage-des-entites-nommees-pour/)

[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/3/) / Repérage des entités nommées pour l'arabe : adaptation non-supervisée et combinaison de systèmes (Named Entity Recognition for Arabic : Unsupervised adaptation and Systems combination) \[in French\]

# Repérage des entités nommées pour l'arabe : adaptation non-supervisée et combinaison de systèmes (Named Entity Recognition for Arabic : Unsupervised adaptation and Systems combination) \[in French\]

**Authors:** [Souhir Gahbiche-Braham](https://md.research.elamur.ai/free/en/authors/souhir-gahbiche-braham/), [Hélène Bonneau-Maynard](https://md.research.elamur.ai/free/en/authors/helene-bonneau-maynard/), [Thomas Lavergne](https://md.research.elamur.ai/free/en/authors/thomas-lavergne/) and [François Yvon](https://md.research.elamur.ai/free/en/authors/francois-yvon/) · **Year:** 2012 · **Venue:** [Proceedings of the Joint Conference JEP-TALN-RECITAL 2012, volume 2: TALN](https://md.research.elamur.ai/free/en/venues/jeptalnrecital/) ([JEP-TALN-RECITAL (the French-speaking conferences on speech and natural language processing) 2012](https://md.research.elamur.ai/free/en/venues/jeptalnrecital/2012/)) · **Publisher:** [ATALA](https://md.research.elamur.ai/free/en/publishers/atala/) · **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/F12-2044/) · [PDF](https://aclanthology.org/F12-2044.pdf)

## Summary
The paper builds a statistical named entity recognizer for Arabic. The goal is to use it as preprocessing for Arabic→French/English machine translation, where unknown names are often copied untranslated. The authors' earlier work found that roughly a quarter of the ~1% unknown forms were named entities. Arabic is hard for this task because it has no capitalization, no written short vowels, many spelling variants of proper names and few lexical resources. Using the Wapiti CRF toolkit with L1-based feature selection, the authors train on the ANER news corpus. They add feature families step by step: word n-grams, affixes, POS tags from a tagger trained on the Arabic Tree Bank, punctuation/digit cues and gazetteers. Their full baseline reaches an overall F-measure of 80.46 under 10-fold cross-validation, with recall notably above an earlier CRF system. They then try to adapt the model to a new domain with no supervision and to combine it with a rule-based system. These steps target the problem that training and test distributions diverge for names over time and topic.

## Key points
- The ANER corpus has 150K+ tokens in 4,871 sentences, annotated in IOB-2 with 9 tags. The entity mix is LOC 40%, PERS 32%, ORG 18% and MISC 10%, and MISC is left out of the scoring.
- Input is transliterated text with no morphological analysis. Evaluation uses the CoNLL 2002 scorer and 10-fold cross-validation, with test folds of about 25,000 words.
- From about 80 million candidate features, Wapiti's L1 penalty keeps around 275,000 active ones.
- The baseline scores P 85.14 / R 76.27 / F 80.46 overall, against F 76.28 for Benajiba & Rosso (2008). The gain comes mostly from recall, and ORG is the weakest class at F 68.75.
- Gazetteers come from ANERGazet, Wikipedia and JRC-Names, giving 3,798 locations, 386 organizations and 13,648 person names. Lookup is done both on the raw word and after stripping prefixes.

## Interesting ideas
- The method generates a very large pool of conjunctive features and lets an L1 penalty decide which ones survive, instead of choosing them by hand.
- Gazetteer matches are tested both on the bare form and on the form with attached prefixes removed, so names glued to clitics still hit the list.
- Feature families are added one at a time, and each step reports precision, recall, F-measure and the number of active features. This shows where gains come from and what they cost in model size.
- The authors justify the module with a downstream count, measuring how many unknown words in translation are named entities.
- The future work proposes training the CRF from partial pre-labels produced by a symbolic system, using latent-variable models.

## For Mova

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

- **I1** (tokens): the tagger and morphology models could start from a wide template space and use sparsity to keep the model small and its active features inspectable.
- **I2** (world): when linking mentions to Wikidata/WordNet entities, a second lookup after stripping possessive 's, hyphen parts or determiners would raise recall in a controlled, explainable way.
- **I3** (learn): ablation reports for parser or frame features could track model size next to accuracy, which would flag additions that add bulk without benefit.
- **I4** (back): translation through the meta-world could log which untranslatable leftovers are names, and route those to copy or transliteration instead of lexical translation.
- **I5** (learn): silver data where witness parsers agree only on some spans could be used as partial supervision rather than forced into full labels.

*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/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works

🧭 Topics (3): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 3,066 works · [🏷️ Named entity recognition](https://md.research.elamur.ai/free/en/topics/named-entity-recognition/) · 214 works · [French](https://md.research.elamur.ai/free/en/topics/french/) · 83 works

## Cite

Cite as: Souhir Gahbiche-Braham, Hélène Bonneau-Maynard, Thomas Lavergne and François Yvon. 2012. Repérage des entités nommées pour l'arabe : adaptation non-supervisée et combinaison de systèmes (Named Entity Recognition for Arabic : Unsupervised adaptation and Systems combination) \[in French\]. Proceedings of the Joint Conference JEP-TALN-RECITAL 2012, volume 2: TALN.

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

```bibtex
@inproceedings{gahbichebraham2012reperage,
  title = {{Repérage des entités nommées pour l'arabe : adaptation non-supervisée et combinaison de systèmes (Named Entity Recognition for Arabic : Unsupervised adaptation and Systems combination) [in French]}},
  author = {Souhir Gahbiche-Braham, Hélène Bonneau-Maynard, Thomas Lavergne and François Yvon},
  booktitle = {Proceedings of the Joint Conference JEP-TALN-RECITAL 2012, volume 2: TALN},
  year = {2012},
  url = {https://aclanthology.org/F12-2044/}
}
```

- [← previous Reflections on Abu Rumi’s Amharic Translation of the Book of Ruth](https://md.research.elamur.ai/free/en/papers/mulugetta-2012-reflections-abu-rumis-amharic-translation/)
- [next → Rescoring N-Best Hypotheses for Arabic Speech Recognition: A Syntax- Mining Approach](https://md.research.elamur.ai/free/en/papers/abuzeina-2012-rescoring-n-best-hypotheses-arabic/)
- [↑ back to papers of 2012](https://md.research.elamur.ai/free/en/papers/2012/page/3/)

## References

26 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. A BDUL H AMID , A. et D ARWISH , K. (2010). [Simplified feature set for Arabic named entity recognition](https://md.research.elamur.ai/free/en/papers/hamid-2010-simplified-feature-set-arabic-named/). In Proc. of the 2010 Named Entities Workshop, pages 110–115, Uppsala.
2. B ÉCHET , F., S AGOT , B. et S TERN , R. (2011). Coopération de méthodes statistiques et symboliques pour l’adaptation non-supervisée d’un système d’étiquetage en entités nommées. In actes de la conférence TALN, Montpellier, France. [find](https://scholar.google.com/scholar?q=B+%C3%89CHET+%2C+F%2E%2C+S+AGOT+%2C+B%2E+et+S+TERN+%2C+R%2E+%282011%29%2E+Coop%C3%A9ration+de+m%C3%A9thodes+statistiques+et+symboliques+pour+l%E2%80%99adaptation+non%2Dsupervis%C3%A9e+d%E2%80%99un+syst%C3%A8me+d%E2%80%99%C3%A9tiquetage+en+entit%C3%A9s+nomm%C3%A9es%2E+In+actes+de+la+conf%C3%A9)
3. B ELLAGARDA , J. R. (2001). An overview of statistical language model adaptation. In Proc. of the ISCA Workshop on Adaptation Methods for Speech Recognition, pages 165–174, Sophia Antipolis. [find](https://scholar.google.com/scholar?q=B+ELLAGARDA+%2C+J%2E+R%2E+%282001%29%2E+An+overview+of+statistical+language+model+adaptation%2E+In+Proc%2E+of+the+ISCA+Workshop+on+Adaptation+Methods+for+Speech+Recognition%2C+pages+165%E2%80%93174%2C+Sophia+Antipolis%2E)
4. B ENAJIBA , Y., D IAB , M. et R OSSO , P. (2008). [Arabic named entity recognition using optimized feature sets](https://md.research.elamur.ai/free/en/papers/benajiba-2008-arabic-named-entity-recognition-optimized/). In Proc. of EMNLP, EMNLP, pages 284–293.
5. B ENAJIBA , Y. et R OSSO , P. (2007). Conquering the NER task for the Arabic language by combining the maximum entropy with POS-tag information. In Proceedings of Workshop on Natural Language-Independant Engineering, IJCAI. [find](https://scholar.google.com/scholar?q=B+ENAJIBA+%2C+Y%2E+et+R+OSSO+%2C+P%2E+%282007%29%2E+Conquering+the+NER+task+for+the+Arabic+language+by+combining+the+maximum+entropy+with+POS%2Dtag+information%2E+In+Proceedings+of+Workshop+on+Natural+Language%2DIndepend)
6. B ENAJIBA , Y. et R OSSO , P. (2008). Arabic named entity recognition using conditional random fields. In Proceedings of the Conference on Language Resources and Evaluation. [find](https://scholar.google.com/scholar?q=B+ENAJIBA+%2C+Y%2E+et+R+OSSO+%2C+P%2E+%282008%29%2E+Arabic+named+entity+recognition+using+conditional+random+fields%2E+In+Proceedings+of+the+Conference+on+Language+Resources+and+Evaluation%2E)
7. B ENAJIBA , Y., R OSSO , P. et B ENEDÍ , J.-M. (2007). Anersys : An arabic named entity recognition system based on maximum entropy. In CICLing, pages 143–153. [find](https://scholar.google.com/scholar?q=B+ENAJIBA+%2C+Y%2E%2C+R+OSSO+%2C+P%2E+et+B+ENED%C3%8D+%2C+J%2E%2DM%2E+%282007%29%2E+Anersys+%3A+An+arabic+named+entity+recognition+system+based+on+maximum+entropy%2E+In+CICLing%2C+pages+143%E2%80%93153%2E)
8. D AUME III, H. (2007). [Frustratingly easy domain adaptation](https://md.research.elamur.ai/free/en/papers/iii-2007-frustratingly-easy-domain-adaptation/). In Proc. of the 45th Annual Meeting of the ACL, pages 256–263, Prague, Czech Republic.
9. D AUME III, H., D EOSKAR , T., M C C LOSKY , D., P LANK , B. et T IEDEMANN , J., éditeurs (2010). Proc. of the 2010 Workshop on Domain Adaptation for NLP. Uppsala, Sweden. [find](https://scholar.google.com/scholar?q=D+AUME+III%2C+H%2E%2C+D+EOSKAR+%2C+T%2E%2C+M+C+C+LOSKY+%2C+D%2E%2C+P+LANK+%2C+B%2E+et+T+IEDEMANN+%2C+J%2E%2C+%C3%A9diteurs+%282010%29%2E+Proc%2E+of+the+2010+Workshop+on+Domain+Adaptation+for+NLP%2E+Uppsala%2C+Sweden%2E)
10. D AUMÉ III, H. et J AGARLAMUDI , J. (2011). [Domain adaptation for machine translation by mining unseen words](https://md.research.elamur.ai/free/en/papers/iii-2011-domain-adaptation-machine-translation-mining/). In ACL, Portland, OR.
11. G AHBICHE -B RAHAM , S., B ONNEAU -M AYNARD , H., L AVERGNE , T. et Y VON , F. (2012). [Joint segmentation and POS tagging for arabic using a CRF-based classifier](https://md.research.elamur.ai/free/en/papers/braham-2012-joint-segmentation-pos-tagging-arabic/). In Proc. of LREC’12.
12. G AHBICHE -B RAHAM , S., B ONNEAU -M AYNARD , H. et Y VON , F. (2011). Two ways to use a noisy parallel news corpus for improving statistical machine translation. In Proc. of Workshop on Building and Using Comparable Corpora, pages 44–51, Portland, OR. [find](https://scholar.google.com/scholar?q=G+AHBICHE+%2DB+RAHAM+%2C+S%2E%2C+B+ONNEAU+%2DM+AYNARD+%2C+H%2E+et+Y+VON+%2C+F%2E+%282011%29%2E+Two+ways+to+use+a+noisy+parallel+news+corpus+for+improving+statistical+machine+translation%2E+In+Proc%2E+of+Workshop+on+Building+and+)
13. G UILLEMIN -L ANNE , S., D EBILI , F., T AHAR , Z. B. et G ACI , C. (2007). Reconnaissance des entités nommées en arabe. In Colloque VSST, Veille Stratégique Scientifique et Technologique. [find](https://scholar.google.com/scholar?q=G+UILLEMIN+%2DL+ANNE+%2C+S%2E%2C+D+EBILI+%2C+F%2E%2C+T+AHAR+%2C+Z%2E+B%2E+et+G+ACI+%2C+C%2E+%282007%29%2E+Reconnaissance+des+entit%C3%A9s+nomm%C3%A9es+en+arabe%2E+In+Colloque+VSST%2C+Veille+Strat%C3%A9gique+Scientifique+et+Technologique%2E)
14. H ABASH , N. (2010). Introduction to Arabic Natural Language Processing. Morgan Claypool. [find](https://scholar.google.com/scholar?q=H+ABASH+%2C+N%2E+%282010%29%2E+Introduction+to+Arabic+Natural+Language+Processing%2E+Morgan+Claypool%2E)
15. H ERMJAKOB , U., K NIGHT , K. et D AUMÉ III, H. (2008). Name translation in statistical machine translation - learning when to transliterate. In Proc. of ACL-08 : HLT, pages 389–397, Ohio. [find](https://scholar.google.com/scholar?q=H+ERMJAKOB+%2C+U%2E%2C+K+NIGHT+%2C+K%2E+et+D+AUM%C3%89+III%2C+H%2E+%282008%29%2E+Name+translation+in+statistical+machine+translation+%2D+learning+when+to+transliterate%2E+In+Proc%2E+of+ACL%2D08+%3A+HLT%2C+pages+389%E2%80%93397%2C+Ohio%2E)
16. J IANG , J. et Z HAI , C. (2007). [Instance weighting for domain adaptation in nlp](https://md.research.elamur.ai/free/en/papers/jiang-2007-instance-weighting-domain-adaptation-nlp/). In Proc. of the 45th Annual Meeting of the ACL, pages 264–271, Prague, Czech Republic.
17. L AFFERTY , J., M C C ALLUM , A. et P EREIRA , F. (2001). Conditional random fields : Probabilistic models for segmenting and labeling sequence data. In Proc. ICML, pages 282–289, San Francisco. [find](https://scholar.google.com/scholar?q=L+AFFERTY+%2C+J%2E%2C+M+C+C+ALLUM+%2C+A%2E+et+P+EREIRA+%2C+F%2E+%282001%29%2E+Conditional+random+fields+%3A+Probabilistic+models+for+segmenting+and+labeling+sequence+data%2E+In+Proc%2E+ICML%2C+pages+282%E2%80%93289%2C+San+Francisco%2E)
18. L AVERGNE , T., C APPÉ , O. et Y VON , F. (2010). [Practical very large scale CRFs](https://md.research.elamur.ai/free/en/papers/lavergne-2010-practical-very-large-scale-crfs/). In Proceedings the 48th Annual Meeting of the Association for Computational Linguistics, pages 504–513.
19. M ALONEY , J. et N IV , M. (1998). [TAGARAB : a fast, accurate Arabic name recognizer using high-precision morphological analysis](https://md.research.elamur.ai/free/en/papers/maloney-1998-tagarab-fast-accurate-arabic-name/). In Proc. of the Workshop on Computational Approaches to Semitic Languages, Semitic ’98, pages 8–15, Stroudsburg, PA, USA.
20. M IHALCEA , R. (2004). Co-training and self-training for word sense disambiguation. In N G , H. T. et R ILOFF , E., éditeurs : HLT-NAACL Workshop : CoNLL-2004, pages 33–40, Boston. [find](https://scholar.google.com/scholar?q=M+IHALCEA+%2C+R%2E+%282004%29%2E+Co%2Dtraining+and+self%2Dtraining+for+word+sense+disambiguation%2E+In+N+G+%2C+H%2E+T%2E+et+R+ILOFF+%2C+E%2E%2C+%C3%A9diteurs+%3A+HLT%2DNAACL+Workshop+%3A+CoNLL%2D2004%2C+pages+33%E2%80%9340%2C+Boston%2E)
21. S AMY , D., M ORENO , A. et M A G UIRAO , J. (2005). A proposal for an Arabic named entity tagger leveraging a parallel corpus. RANLP ’05. [find](https://scholar.google.com/scholar?q=S+AMY+%2C+D%2E%2C+M+ORENO+%2C+A%2E+et+M+A+G+UIRAO+%2C+J%2E+%282005%29%2E+A+proposal+for+an+Arabic+named+entity+tagger+leveraging+a+parallel+corpus%2E+RANLP+%E2%80%9905%2E)
22. S HAALAN , K. et R AZA , H. (2009). NERA : Named entity recognition for arabic. Journal of the American Society for Information Science and Technology, 60(9):1652–1663. [find](https://scholar.google.com/scholar?q=S+HAALAN+%2C+K%2E+et+R+AZA+%2C+H%2E+%282009%29%2E+NERA+%3A+Named+entity+recognition+for+arabic%2E+Journal+of+the+American+Society+for+Information+Science+and+Technology%2C+60%289%29%3A1652%E2%80%931663%2E)
23. S OKOLOVSKA , N., C APPÉ , O. et Y VON , F. (2009). Sélection de caractéristiques pour les champs aléatoires conditionnels par pénalisation l 1 . TAL, 50(3):139–171. [find](https://scholar.google.com/scholar?q=S+OKOLOVSKA+%2C+N%2E%2C+C+APP%C3%89+%2C+O%2E+et+Y+VON+%2C+F%2E+%282009%29%2E+S%C3%A9lection+de+caract%C3%A9ristiques+pour+les+champs+al%C3%A9atoires+conditionnels+par+p%C3%A9nalisation+l+1+%2E+TAL%2C+50%283%29%3A139%E2%80%93171%2E)
24. Z AGHOUANI , W., P OULIQUEN , B., E BRAHIM , M. et S TEINBERGER , R. (2010). [Adapting a resource-light highly multilingual named entity recognition system to arabic](https://md.research.elamur.ai/free/en/papers/zaghouani-2010-adapting-resource-light-highly-multilingual/). Proceedings of the Seventh conference on International Language Resources and Evaluation (LREC’10), pages 563–567.
25. Z HANG , M., L I , H., K UMARAN , A. et L IU , M. (2011). Report of news2011 machin transliteration shared task. In Proceedings of the 2011 Named Entities Workshop. [find](https://scholar.google.com/scholar?q=Z+HANG+%2C+M%2E%2C+L+I+%2C+H%2E%2C+K+UMARAN+%2C+A%2E+et+L+IU+%2C+M%2E+%282011%29%2E+Report+of+news2011+machin+transliteration+shared+task%2E+In+Proceedings+of+the+2011+Named+Entities+Workshop%2E)
26. Z ITOUNI , I., S ORENSEN , J., L UO , X. et F LORIAN , R. (2005). [The impact of morphological stemming on Arabic mention detection and coreference resolution](https://md.research.elamur.ai/free/en/papers/zitouni-2005-impact-morphological-stemming-arabic-mention/). In Proc. of Workshop on Computational Approaches to Semitic Languages, pages 63–70, Ann Arbor, Michigan.

*Updated 2 November 2026*