---
title: "NAKBA NLP 2026 — paper by Hamoud et al. (2026)"
description: "Subtask 2 asked teams to build automatic HTR systems, scored with CER and WER. 22 references, cited by 5 papers here."
authors: ["Hadi Hamoud", "Ahmad Ali Chamseddine", "Bilal Shalash", "Firas Ben Abid", "Mustafa Jarrar", "Chadi Abou Chakra", "Bernard Ghanem", "Fadi A. Zaraket"]
year: 2026
venue: "Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026"
source: "https://aclanthology.org/2026.nakbanlp-1.7/"
doi: "10.63317/3iakmct86er7"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/hamoud-2026-nakba-nlp-2026-shared-task/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/hamoud-2026-nakba-nlp-2026-shared-task/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/3/) / NAKBA NLP 2026: Shared Task on Arabic Handwritten Manuscript Understanding (Palestine Memory–Omar Al-Saleh Memoir)

# NAKBA NLP 2026: Shared Task on Arabic Handwritten Manuscript Understanding (Palestine Memory–Omar Al-Saleh Memoir)

**Authors:** [Hadi Hamoud](https://md.research.elamur.ai/free/en/authors/hadi-hamoud/), Ahmad Ali Chamseddine, [Bilal Shalash](https://md.research.elamur.ai/free/en/authors/bilal-shalash/), Firas Ben Abid, [Mustafa Jarrar](https://md.research.elamur.ai/free/en/authors/mustafa-jarrar/), [Chadi Abou Chakra](https://md.research.elamur.ai/free/en/authors/chadi-abou-chakra/), [Bernard Ghanem](https://md.research.elamur.ai/free/en/authors/bernard-ghanem/) and [Fadi A. Zaraket](https://md.research.elamur.ai/free/en/authors/fadi-zaraket/) · **Year:** 2026 · **Venue:** [Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026](https://md.research.elamur.ai/free/en/venues/nakbanlp/) ([International Workshop on Nakba Narratives as Language Resources @ LREC 2026](https://md.research.elamur.ai/free/en/venues/nakbanlp/2026/)) · **Publisher:** [European Language Resources Association (ELRA)](https://md.research.elamur.ai/free/en/publishers/european-language-resources-association-elra/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.nakbanlp-1.7/) · [PDF](https://aclanthology.org/2026.nakbanlp-1.7.pdf) · [DOI 10.63317/3iakmct86er7](https://doi.org/10.63317/3iakmct86er7) · [OpenAlex](https://openalex.org/W7166582008)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · [5](https://md.research.elamur.ai/free/en/papers/hamoud-2026-nakba-nlp-2026-shared-task/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Historical Arabic handwriting remains hard to convert into searchable text. The causes include variable letterforms, damaged pages and Arabic-specific issues such as dropped diacritics. Open benchmarks for this problem are also scarce. The NAKBA NLP 2026 shared task addresses the gap with one large single-author corpus: the Omar Al-Saleh Memoir Collection from the Palestine Memory Project, covering notebooks written 1951–1965. The task has two tracks.

- **Subtask 1** asked teams to hand-transcribe unlabelled line images. Their work was scored against hidden gold lines.
- **Subtask 2** asked teams to build automatic HTR systems, scored with CER and WER.

Results:
- Three human-transcription teams reached CER between 0.06 and 0.11, so even people find this material difficult.
- All seven automatic systems beat the organisers' Qwen-VL-based baseline.
- Misraj AI was best at corpus level.
- Ketaba OCR led when errors were averaged per line.

The data, scoring scripts and outputs are released under CC-BY-4.0. This gives Arabic manuscript HTR a reusable reference point that it previously lacked.

## Key points
- The corpus has 6,395 scanned pages and 1,597,025 words. It contains 50,685 sentences and 161,394 unique word types, with a type–token ratio of 0.1011. It is split into per-year notebooks, with "b" suffixes marking second notebooks from the same year.
- The organiser baseline scored CER 0.368 and WER 0.691. Misraj AI scored CER 0.079 and WER 0.244, a relative CER drop of 78.5%. Ketaba OCR had the lowest per-line CER, at 0.082.
- In Subtask 1, three teams achieved CER of 0.061–0.114. Generative AI was not allowed for manual transcription. Because each team worked on a different batch, inter-annotator agreement could not be computed.
- The error analysis attributes errors to four main causes:
  - word-boundary mistakes: about 24%
  - missing or shifted dots and diacritics: 18%
  - degraded regions, mostly in the 1951–1953 notebooks: 12%
  - rare words and proper nouns: 8%
- Stated limitations: the corpus reflects only one writer, and evaluation is line-level only.

## Interesting ideas
- Scoring the same outputs both pooled over the corpus and averaged per line changed the winner, which shows that each aggregation favours a different kind of system.
- Human contributions were graded by hiding gold-labelled lines among the unlabelled work, so annotator quality was measured without a separate test session.
- Several teams used the same Qwen-VL family as the baseline yet beat it widely, which points to augmentation, training recipe and prompting rather than architecture as the main lever.
- A small error taxonomy with percentage shares, covering boundaries, dots, degradation and rare names, turns one opaque error rate into actionable targets.
- Rare names fail because language-model priors cannot resolve visual ambiguity for infrequent items, so names need support beyond the general model.

## For Mova

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

- **I1** (learn): Report LAS both micro-averaged over tokens and macro-averaged per sentence for every parser run, so a model that only wins on long sentences is not mistaken for an overall improvement.
- **I2** (learn): Slip known gold sentences into batches sent to LLM teacher tiers or witness parsers, and use their accuracy on those items to weight the silver data they produce.
- **I3** (tree): Before redesigning gparse, run ablations on training data mix and hyperparameters, since recipe changes may yield larger gains than a new architecture.
- **I4** (tokens): Classify tokenizer and tagger errors into similar buckets, such as split or merge errors, clitics and rare proper nouns, and track each bucket's share over time.
- **I5** (world): Back up entity recognition with the Wikidata/WordNet knowledge layer, so unseen proper nouns can still be resolved against known entities rather than guessed from context alone.

*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

- [Hamad bin Khalifa University](https://md.research.elamur.ai/free/en/universities/hamad-bin-khalifa-university/) — University, Doha, Qatar · 141 works · 14,175 citations
- [King Abdullah University of Science and Technology](https://md.research.elamur.ai/free/en/universities/king-abdullah-university-of-science-and-technology/) — University, Jeddah, Saudi Arabia · 27 works · 1,084 citations
- [Arab Center for Research and Policy Studies](https://md.research.elamur.ai/free/en/labs/arab-center-for-research-and-policy-studies/) — Research institute, Doha, Qatar · 8 works · 69 citations
- [Doha Institute for Graduate Studies](https://md.research.elamur.ai/free/en/universities/doha-institute-for-graduate-studies/) — University, Doha, Qatar · 2 works · 32 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works · [📚 Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) · 519 works · [⚙️ Engineering](https://md.research.elamur.ai/free/en/shelves/engineering/) · 133 works · [⚙️ Media Technology](https://md.research.elamur.ai/free/en/shelves/engineering/media-technology/) · 34 works

🧭 Topics (1): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works

## Cite

Cite as: Hadi Hamoud, Ahmad Ali Chamseddine, Bilal Shalash, Firas Ben Abid, Mustafa Jarrar, Chadi Abou Chakra, Bernard Ghanem and Fadi A. Zaraket. 2026. NAKBA NLP 2026: Shared Task on Arabic Handwritten Manuscript Understanding (Palestine Memory–Omar Al-Saleh Memoir). Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026.

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

```bibtex
@inproceedings{hamoud2026nakba,
  title = {{NAKBA NLP 2026: Shared Task on Arabic Handwritten Manuscript Understanding (Palestine Memory–Omar Al-Saleh Memoir)}},
  author = {Hadi Hamoud, Ahmad Ali Chamseddine, Bilal Shalash, Firas Ben Abid, Mustafa Jarrar, Chadi Abou Chakra, Bernard Ghanem and Fadi A. Zaraket},
  booktitle = {Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026},
  year = {2026},
  doi = {10.63317/3iakmct86er7},
  url = {https://doi.org/10.63317/3iakmct86er7}
}
```

- [← previous MulVul: Retrieval-augmented Multi-Agent Code Vulnerability Detection via Cross-Model Prompt Evolution](https://md.research.elamur.ai/free/en/papers/wu-2026-mulvul-retrieval-augmented-multi-agent/)
- [next → OctoTools: A Multi-Agent Framework with Extensible Tools for Complex Reasoning](https://md.research.elamur.ai/free/en/papers/lu-2026-octotools-multi-agent-framework-extensible/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/3/)

## References

22 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. O. S. Al-Barghouti and K. Totah. 1923. Tarikh Filastin [History of Palestine]. Bayt al-Maqdis Press, Jerusalem. [find](https://scholar.google.com/scholar?q=O%2E+S%2E+Al%2DBarghouti+and+K%2E+Totah%2E+1923%2E+Tarikh+Filastin+%5BHistory+of+Palestine%5D%2E+Bayt+al%2DMaqdis+Press%2C+Jerusalem%2E)
2. N. Alrobah and S. Alzahrani. 2022. Arabic handwritten text extraction: A survey. IEEE Access, 10:50437–50465. [find](https://scholar.google.com/scholar?q=N%2E+Alrobah+and+S%2E+Alzahrani%2E+2022%2E+Arabic+handwritten+text+extraction%3A+A+survey%2E+IEEE+Access%2C+10%3A50437%E2%80%9350465%2E)
3. Arab Center for Research and Policy Studies. 2024. Palestine Memory Project. [find](https://scholar.google.com/scholar?q=Arab+Center+for+Research+and+Policy+Studies%2E+2024%2E+Palestine+Memory+Project%2E)
4. J. Bai, S. Bai, S. Yang, et al. 2023. [Qwen-VL: A versatile vision-language model for understanding, localization, text reading, and beyond](https://md.research.elamur.ai/free/en/papers/bai-2023-qwen-vl-versatile-vision-language/). arXiv preprint arXiv:2308.12966. [↗](https://arxiv.org/abs/2308.12966)
5. Z. Chen et al. 2024. [InternVL: Scaling Up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://md.research.elamur.ai/free/en/papers/chen-2023-internvl-scaling-up-vision/). arXiv preprint arXiv:2312.14238. [↗](https://arxiv.org/abs/2312.14238)
6. C. Cieri, M. Maamouri, S. Strassel, et al. 2016. MADCAT: Multilingual automatic document classification, analysis, and translation corpus. [find](https://scholar.google.com/scholar?q=C%2E+Cieri%2C+M%2E+Maamouri%2C+S%2E+Strassel%2C+et+al%2E+2016%2E+MADCAT%3A+Multilingual+automatic+document+classification%2C+analysis%2C+and+translation+corpus%2E)
7. C. Clausner, A. Antonacopoulos, N. McGregor, and D. Wilson-Nunn. 2018. ICFHR 2018 competition on recognition of historical Arabic scientific manuscripts – RASM2018. In Proc. ICFHR, pages 471–476. [find](https://scholar.google.com/scholar?q=C%2E+Clausner%2C+A%2E+Antonacopoulos%2C+N%2E+McGregor%2C+and+D%2E+Wilson%2DNunn%2E+2018%2E+ICFHR+2018+competition+on+recognition+of+historical+Arabic+scientific+manuscripts+%E2%80%93+RASM2018%2E+In+Proc%2E+ICFHR%2C+pages+471%E2%80%93476%2E)
8. A. Graves, S. Fernández, F. Gomez, and J. Schmidhuber. 2006. Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks. In Proceedings of ICML, pages 369–376. [find](https://scholar.google.com/scholar?q=A%2E+Graves%2C+S%2E+Fern%C3%A1ndez%2C+F%2E+Gomez%2C+and+J%2E+Schmidhuber%2E+2006%2E+Connectionist+temporal+classification%3A+Labelling+unsegmented+sequence+data+with+recurrent+neural+networks%2E+In+Proceedings+of+ICML%2C+pages+36)
9. Institute for Palestine Studies. a. Omar al-saleh al-barghouti (1894–1965). Accessed: 2025. [find](https://scholar.google.com/scholar?q=Institute+for+Palestine+Studies%2E+a%2E+Omar+al%2Dsaleh+al%2Dbarghouti+%281894%E2%80%931965%29%2E+Accessed%3A+2025%2E)
10. Institute for Palestine Studies. b. Palestine Social History Archives: Omar Saleh Al-Barghouti Diaries. Accessed: 2025. [find](https://scholar.google.com/scholar?q=Institute+for+Palestine+Studies%2E+b%2E+Palestine+Social+History+Archives%3A+Omar+Saleh+Al%2DBarghouti+Diaries%2E+Accessed%3A+2025%2E)
11. Mustafa Jarrar, Mo El-Haj, Amal Haddad, Serin Atiani, Shadi Abudalfa, Khalil Sima’an, Paul Rayson, and Camille Mansour, editors. 2026. Proceedings of the second International Workshop on Nakba Narratives as Language Resources. Association for Computational Linguistics, Spain. [find](https://scholar.google.com/scholar?q=Mustafa+Jarrar%2C+Mo+El%2DHaj%2C+Amal+Haddad%2C+Serin+Atiani%2C+Shadi+Abudalfa%2C+Khalil+Sima%E2%80%99an%2C+Paul+Rayson%2C+and+Camille+Mansour%2C+editors%2E+2026%2E+Proceedings+of+the+second+International+Workshop+on+Nakba+Narrati)
12. G. Jocher, A. Chaurasia, and J. Qiu. 2024. Ultralytics YOLO. Version 8.3.0+. [find](https://scholar.google.com/scholar?q=G%2E+Jocher%2C+A%2E+Chaurasia%2C+and+J%2E+Qiu%2E+2024%2E+Ultralytics+YOLO%2E+Version+8%2E3%2E0%2B%2E)
13. M. Kassis, A. Abdalhaleem, A. Droby, R. Alaasam, and J. El-Sana. 2017. VML-HD: The historical document dataset. In Proc. ICDAR Workshops. [find](https://scholar.google.com/scholar?q=M%2E+Kassis%2C+A%2E+Abdalhaleem%2C+A%2E+Droby%2C+R%2E+Alaasam%2C+and+J%2E+El%2DSana%2E+2017%2E+VML%2DHD%3A+The+historical+document+dataset%2E+In+Proc%2E+ICDAR+Workshops%2E)
14. R. Khanam and M. Hussain. 2024. YOLOv11: An Overview of the Key Architectural Enhancements. arXiv preprint arXiv:2410.17725. [↗](https://arxiv.org/abs/2410.17725)
15. M. Li, T. Lv, J. Chen, et al. 2023. TrOCR: Transformer-based optical character recognition with pre-trained models. [find](https://scholar.google.com/scholar?q=M%2E+Li%2C+T%2E+Lv%2C+J%2E+Chen%2C+et+al%2E+2023%2E+TrOCR%3A+Transformer%2Dbased+optical+character+recognition+with+pre%2Dtrained+models%2E)
16. S. A. Mahmoud, I. Ahmad, W. G. Al-Khatib, et al. 2014. KHATT: An open Arabic offline handwritten text database. Pattern Recognition, 47(3):1004– 1011. [↗](https://doi.org/10.1016/j.patcog.2013.08.009)
17. G. Muehlberger, L. Seaward, M. Terras, et al. 2019. Transforming scholarship in the archives through handwritten text recognition: Transkribus as a case study. Journal of Documentation, 75(5):954–976. [↗](https://doi.org/10.7892/boris.137985)
18. M. Pechwitz, S. S. Maddouri, V. Märgner, et al. 2002. IFN/ENIT – database of handwritten Arabic words. In Proc. CIFED, pages 129–136. [find](https://scholar.google.com/scholar?q=M%2E+Pechwitz%2C+S%2E+S%2E+Maddouri%2C+V%2E+M%C3%A4rgner%2C+et+al%2E+2002%2E+IFN%2FENIT+%E2%80%93+database+of+handwritten+Arabic+words%2E+In+Proc%2E+CIFED%2C+pages+129%E2%80%93136%2E)
19. M. Romanov et al. 2023. OpenITI: Open Islamicate Texts Initiative. [find](https://scholar.google.com/scholar?q=M%2E+Romanov+et+al%2E+2023%2E+OpenITI%3A+Open+Islamicate+Texts+Initiative%2E)
20. V. Romero, J. A. Sánchez, A. H. Toselli, et al. 2012. The ESPOSALLES database: An ancient marriage license corpus for off-line handwriting recognition. Pattern Recognition, 46(6):1658– 1669. [↗](https://doi.org/10.1016/j.patcog.2012.11.024)
21. A. Saeed, C. Ascherl, N. Uchigasaki, et al. 2024. Muharaf: Manuscripts of handwritten Arabic dataset for cursive text recognition. In NeurIPS Datasets and Benchmarks Track. [find](https://scholar.google.com/scholar?q=A%2E+Saeed%2C+C%2E+Ascherl%2C+N%2E+Uchigasaki%2C+et+al%2E+2024%2E+Muharaf%3A+Manuscripts+of+handwritten+Arabic+dataset+for+cursive+text+recognition%2E+In+NeurIPS+Datasets+and+Benchmarks+Track%2E)
22. R. Xu, Y. Wang, S. Zhang, et al. 2024. Florence-2: Advancing a unified representation for a variety of vision tasks. arXiv preprint arXiv:2311.06242. [↗](https://arxiv.org/abs/2311.06242)

*Updated 1 November 2026*