---
title: "Arabic Diacritization with Recurrent Neural Networks — Belinkov 2015"
description: "The paper asks whether Arabic diacritics (short vowels, case endings, gemination) can be restored from diacritized… 14 references, cited by 34 papers here."
authors: ["Yonatan Belinkov", "James Glass"]
year: 2015
venue: "Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing"
source: "https://aclanthology.org/D15-1274/"
doi: "10.18653/v1/d15-1274"
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/belinkov-2015-arabic-diacritization-recurrent-neural-networks/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/belinkov-2015-arabic-diacritization-recurrent-neural-networks/)

[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/2/) / Arabic Diacritization with Recurrent Neural Networks

# Arabic Diacritization with Recurrent Neural Networks

**Authors:** [Yonatan Belinkov](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/); [James Glass](https://md.research.elamur.ai/free/en/authors/james-glass/) · **Year:** 2015 · **Venue:** Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing **Link:** [aclanthology.org/D15-1274](https://aclanthology.org/D15-1274/) · library `aclbulk-2015-belinkov-arabic-diacritization-recurrent-neural-networks` · **Licence:** [CC BY-NC-SA 3.0](https://creativecommons.org/licenses/by-nc-sa/3.0/) (ACL Anthology, before 2016) — summary only

[**59** citations](https://openalex.org/works?filter=cites:W2250751111) OpenAlex, 2026-10-14 · [**34** citations in our library](https://md.research.elamur.ai/free/en/papers/belinkov-2015-arabic-diacritization-recurrent-neural-networks/cited-by/) of the 140,884 papers whose bibliographies we read · [Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing](https://aclanthology.org/volumes/D15-1/) · [licence](https://creativecommons.org/licenses/by-nc-sa/3.0/)

- [Publisher's page](https://aclanthology.org/D15-1274/)
- [PDF](https://aclanthology.org/D15-1274.pdf)
- [DOI 10.18653/v1/d15-1274](https://doi.org/10.18653/v1/d15-1274)
- [OpenAlex](https://openalex.org/W2250751111)

**Authors (2)**

1. [Yonatan Belinkov](https://openalex.org/A5051184573) [ORCID](https://orcid.org/0000-0002-6280-5045) · [Semantic Scholar](https://www.semanticscholar.org/author/2083259) · h 43 · 7,752 citations · 202 works
2. [James Glass](https://openalex.org/A5112758056) [ORCID](https://orcid.org/0000-0002-3097-360X) · [Semantic Scholar](https://www.semanticscholar.org/author/145898106) · h 70 · 20,869 citations · 510 works

Citations and author metrics: OpenAlex, 2026-10-14; authors with the same name are not always told apart.

## What it is about

The paper asks whether Arabic diacritics (short vowels, case endings, gemination) can be restored from diacritized text alone, without a morphological analyser or a tagger. The authors treat diacritization as sequence labelling: each letter gets a label for zero, one or two diacritics, and a character-level recurrent network (LSTM, bidirectional, up to three layers) predicts it from a window of five letters. Data is the Penn Arabic Treebank split used by Zitouni and Sarikaya. Deeper bidirectional LSTMs do best on the development set. On the test split the three-layer model beats the lexical MaxEnt baseline by 3.25 points of diacritic error rate (a 40% error reduction) and beats their best model that used a segmenter and tagger. The main errors are confusions among the short vowels a, i, u and the empty label; the proper name Tueini is missed by all models.

## Ideas

- I2. Their error analysis as a standing test: a confusion matrix of predicted against gold vowel classes (a/i/u/none) for the Hebrew vowel signs, reported on the held-out set, to show which vowel classes the model confuses.
- I3. Compose the label from a vowel and a consonant mark (their gemination label is a combination of a diacritic with a marker). For Hebrew, a dagesh-plus-vowel label set could let one tag carry both marks instead of separate decisions.

## For Mova

- I1. A no-network reference for the unpointed mode of `tools/hbo`: a lexical maximum-entropy or count model over a five-letter window (as their MaxEnt-lexical row), trained on the same pointed text, so that any neural or rule-based gain is measured against a cheap, transparent baseline.

I1 and I2 are measures for `tools/hbo` (unpointed mode): the baseline and the confusion table are tests, not new rules. I3 goes into [languages/hbo/03-phonology.md](https://mova.elamur.ai/free/en/docs/languages/hbo/03-phonology/) as a label design for pointed Hebrew. The network itself is not used: Mova's SLM stays rule- and table-based.

## Institutions

[🎓**Massachusetts Institute of Technology** Cambridge · United States 125 works · 60,031 citations](https://md.research.elamur.ai/free/en/universities/massachusetts-institute-of-technology/) · [🔬**MIT Computer Science and Artificial Intelligence Laboratory** Cambridge · United States 45 works · 26,807 citations](https://md.research.elamur.ai/free/en/labs/mit-computer-science-and-artificial-intelligence-laboratory/)

Topics[Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/)

## Cite

Cite as Yonatan Belinkov and James Glass. 2015. Arabic Diacritization with Recurrent Neural Networks. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing.

The paper itself, not this card. BibTeX from the library's record of the work. [DOI 10.18653/v1/d15-1274](https://doi.org/10.18653/v1/d15-1274)

```bibtex
@inproceedings{belinkov2015arabic,
  title = {{Arabic Diacritization with Recurrent Neural Networks}},
  author = {Yonatan Belinkov and James Glass},
  booktitle = {Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing},
  year = {2015},
  doi = {10.18653/v1/d15-1274},
  url = {https://doi.org/10.18653/v1/d15-1274}
}
```

- [← previous Discourse parsing for multi-party chat dialogues](https://md.research.elamur.ai/free/en/papers/afantenos-2015-discourse-parsing-multi-party-chat/)
- [next → Machine Comprehension with Discourse Relations](https://md.research.elamur.ai/free/en/papers/narasimhan-2015-machine-comprehension-discourse-relations/)
- [↑ back to papers of 2015](https://md.research.elamur.ai/free/en/papers/2015/page/2/)

## References

14 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Tuka Al Hanai and James Glass. 2014. Lexical Modeling for Arabic ASR: A Systematic Approach. In Proceedings of INTERSPEECH. [find](https://scholar.google.com/scholar?q=Tuka+Al+Hanai+and+James+Glass%2E+2014%2E+Lexical+Modeling+for+Arabic+ASR%3A+A+Systematic+Approach%2E+In+Proceedings+of+INTERSPEECH%2E)
2. Ahmad Al Sallab, Mohsen Rashwan, Hazem M. Raafat, and Ahmed Rafea. 2014. Automatic Arabic diacritics restoration based on deep nets. In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP). [find](https://scholar.google.com/scholar?q=Ahmad+Al+Sallab%2C+Mohsen+Rashwan%2C+Hazem+M%2E+Raafat%2C+and+Ahmed+Rafea%2E+2014%2E+Automatic+Arabic+diacritics+restoration+based+on+deep+nets%2E+In+Proceedings+of+the+EMNLP+2014+Workshop+on+Arabic+Natural+Languag)
3. Alex Graves, Abdel-rahman Mohamed, and Geoffrey Hinton. 2013. Speech recognition with deep recurrent neural networks. In Proceedings of ICASSP. [find](https://scholar.google.com/scholar?q=Alex+Graves%2C+Abdel%2Drahman+Mohamed%2C+and+Geoffrey+Hinton%2E+2013%2E+Speech+recognition+with+deep+recurrent+neural+networks%2E+In+Proceedings+of+ICASSP%2E)
4. Nizar Habash and Owen Rambow. 2007. Arabic Diacritization through Full Morphological Tagging. In Proceedings of HLT-NAACL. [our card](https://md.research.elamur.ai/free/en/papers/habash-2007-arabic-diacritization-through-full-morphological/)
5. Nizar Habash, Abdelhadi Soudi, and Timothy Buckwalter. 2007. On Arabic Transliteration. In Abdelhadi Soudi, Günter Neumann, and Antal Van den Bosch, editors, Arabic Computational Morphology: Knowledge-based and Empirical Methods, pages 15–22. Springer. [find](https://scholar.google.com/scholar?q=Nizar+Habash%2C+Abdelhadi+Soudi%2C+and+Timothy+Buckwalter%2E+2007%2E+On+Arabic+Transliteration%2E+In+Abdelhadi+Soudi%2C+Gu%CC%88nter+Neumann%2C+and+Antal+Van+den+Bosch%2C+editors%2C+Arabic+Computational+Morphology%3A+Knowledg)
6. 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 Second International Conference on Arabic Language Resources and Tools. [find](https://scholar.google.com/scholar?q=Nizar+Habash%2C+Owen+Rambow%2C+and+Ryan+Roth%2E+2009%2E+MADA%2BTOKAN%3A+A+Toolkit+for+Arabic+Tokenization%2C+Diacritization%2C+Morphological+Disambiguation%2C+POS+Tagging%2C+Stemming+and+Lemmatization%2E+In+Proceedings+of+)
7. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long Short-Term Memory. Neural computation, 9(8):1735–1780. [↗](https://doi.org/10.1162/neco.1997.9.8.1735)
8. Mohamed Maamouri, Ann Bies, and Seth Kulick. 2006. Diacritization: A challenge to Arabic treebank annotation and parsing. In Proceedings of the British Computer Society Arabic NLP/MT Conference. [find](https://scholar.google.com/scholar?q=Mohamed+Maamouri%2C+Ann+Bies%2C+and+Seth+Kulick%2E+2006%2E+Diacritization%3A+A+challenge+to+Arabic+treebank+annotation+and+parsing%2E+In+Proceedings+of+the+British+Computer+Society+Arabic+NLP%2FMT+Conference%2E)
9. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013a. Efficient Estimation of Word Representations in Vector Space. In Proceedings of Workshop at ICLR. [our card](https://md.research.elamur.ai/free/en/papers/mikolov-2013-efficient-estimation-word-representations-vector/)
10. Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013b. Distributed Representations of Words and Phrases and their Compositionality. In Proceedings of NIPS. [our card](https://md.research.elamur.ai/free/en/papers/mikolov-2013-distributed-representations-words-phrases-their/)
11. M.A.A. Rashwan, M.A.S.A.A. Al-Badrashiny, M. Attia, S.M. Abdou, and A. Rafea. 2011. A Stochastic Arabic Diacritizer Based on a Hybrid of Factorized and Unfactorized Textual Features. IEEE Transactions on Audio, Speech, and Language Processing, 19(1):166–175, Jan. [find](https://scholar.google.com/scholar?q=M%2EA%2EA%2E+Rashwan%2C+M%2EA%2ES%2EA%2EA%2E+Al%2DBadrashiny%2C+M%2E+Attia%2C+S%2EM%2E+Abdou%2C+and+A%2E+Rafea%2E+2011%2E+A+Stochastic+Arabic+Diacritizer+Based+on+a+Hybrid+of+Factorized+and+Unfactorized+Textual+Features%2E+IEEE+Transactions)
12. Felix Weninger, Johannes Bergmann, and Björn Schuller. 2015. Introducing CURRENNT: The Munich Open-Source CUDA RecurREnt Neural Network Toolkit. JMLR, 16:547–551. [find](https://scholar.google.com/scholar?q=Felix+Weninger%2C+Johannes+Bergmann%2C+and+Bjo%CC%88rn+Schuller%2E+2015%2E+Introducing+CURRENNT%3A+The+Munich+Open%2DSource+CUDA+RecurREnt+Neural+Network+Toolkit%2E+JMLR%2C+16%3A547%E2%80%93551%2E)
13. Imed Zitouni and Ruhi Sarikaya. 2009. Arabic Diacritic Restoration Approach Based on Maximum Entropy Models. Comput. Speech Lang., 23(3). [find](https://scholar.google.com/scholar?q=Imed+Zitouni+and+Ruhi+Sarikaya%2E+2009%2E+Arabic+Diacritic+Restoration+Approach+Based+on+Maximum+Entropy+Models%2E+Comput%2E+Speech+Lang%2E%2C+23%283%29%2E)
14. Imed Zitouni, Jeffrey S. Sorensen, and Ruhi Sarikaya. 2006. Maximum Entropy Based Restoration of Arabic Diacritics. In Proceedings of COLING and ACL. [our card](https://md.research.elamur.ai/free/en/papers/zitouni-2006-maximum-entropy-based-restoration-arabic/)