---
title: "Building Annotated Written and Spoken Arabic LRs in NEMLAR"
description: "Arabic lacked basic language resources, and this held back both research and industry. 93% accuracy. 14 references, cited by 7 papers here."
authors: ["M. Yaseen", "M. Attia", "B. Maegaard", "K. Choukri", "N. Paulsson", "S. Haamid", "S. Krauwer", "C. Bendahman", "H. Fersøe", "M. Rashwan", "B. Haddad", "C. Mukbel", "A. Mouradi", "A. Al-Kufaishi", "M. Shahin", "N. Chenfour", "A. Ragheb"]
year: 2006
venue: "Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06)"
source: "https://aclanthology.org/L06-1067/"
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/yaseen-2006-building-annotated-written-spoken-arabic/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/yaseen-2006-building-annotated-written-spoken-arabic/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2006](https://md.research.elamur.ai/free/en/papers/2006/page/2/) / Building Annotated Written and Spoken Arabic LRs in NEMLAR Project

# Building Annotated Written and Spoken Arabic LRs in NEMLAR Project

**Authors:** [M. Yaseen](https://md.research.elamur.ai/free/en/authors/mustafa-yaseen/), M. Attia, B. Maegaard, [K. Choukri](https://md.research.elamur.ai/free/en/authors/khalid-choukri/), [N. Paulsson](https://md.research.elamur.ai/free/en/authors/niklas-paulsson/), S. Haamid, [S. Krauwer](https://md.research.elamur.ai/free/en/authors/steven-krauwer/), [C. Bendahman](https://md.research.elamur.ai/free/en/authors/chomicha-bendahman/), H. Fersøe, [M. Rashwan](https://md.research.elamur.ai/free/en/authors/mohsen-rashwan/), B. Haddad, C. Mukbel, A. Mouradi, A. Al-Kufaishi, [M. Shahin](https://md.research.elamur.ai/free/en/authors/mostafa-shahin/), [N. Chenfour](https://md.research.elamur.ai/free/en/authors/n-chenfour/) and A. Ragheb · **Year:** 2006 · **Venue:** [Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06)](https://md.research.elamur.ai/free/en/venues/lrec/) ([LREC 2006](https://md.research.elamur.ai/free/en/venues/lrec/2006/)) · **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/L06-1067/) · [PDF](https://aclanthology.org/L06-1067.pdf) · [OpenAlex](https://openalex.org/W2336597709)

**Citations:** 36 citations (OpenAlex, 2026-10-14) · [7](https://md.research.elamur.ai/free/en/papers/yaseen-2006-building-annotated-written-spoken-arabic/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Arabic lacked basic language resources, and this held back both research and industry. The EC-funded NEMLAR network brought European and Middle Eastern partners together to fix this. They surveyed existing resources and consulted industry. Based on these findings, they specified an Arabic BLARK (Basic Language Resource Kit), the minimum collection of resources required for pre-competitive research, teaching and core NLP components.

From the priority list, the partners built three resources:
- a written corpus of about 500K words of Standard Arabic, annotated for lexical analysis, PoS tags and full vowelization;
- a TTS speech database with 5 hours each from a male and a female speaker;
- 40 hours of transcribed Modern Standard Arabic broadcast news.

The written corpus was pre-annotated automatically with the Fassieh tool, and linguists then corrected the output. ELDA and partner institutions checked all three resources against written validation criteria, and ELRA distributes them. The paper matters because it shows how a standards-driven, budget-aware consortium can produce a reusable starter kit for an under-resourced language.

## Key points
- The written corpus has four parallel versions: raw text, vowelized text, PoS tags, and lexical analysis (prefix, root, pattern, suffix, each with a mnemonic and an ID). All files are in Unicode.
- Fassieh's automatic engine reached about 93% accuracy. Linguists picked from ranked candidate analyses and raised this to 99%+. They accepted the top-ranked analysis in roughly 93% of cases and chose an alternative in about 7%.
- The budget was about 25 man-months, and this fixed the size at 500K words. The text spans late 1990s–2005, and domains are weighted by importance for applications (general news 20%, dictionary explanations 10.4%, political news 10.2%, and so on).
- Text sources were IslamOnLine 75%, RDI internal documents 20%, Annahar 3% and other free sources 2%, all with IPR clearance letters.
- The broadcast news corpus covers 4 sources with 209 male and 50 female speakers. Transcriptions mark speaker turns, topics, named entities and channel information, and come with a phonetic lexicon in Arabic SAMPA.

## Interesting ideas
- Pre-annotate automatically, then have humans choose among ranked candidate analyses instead of typing labels from scratch; most decisions become a single "accept top-1" action, which makes high-accuracy annotation cheap.
- Store every annotation layer (raw, morphological, PoS, phonetic) as a separate parallel file aligned to the same input, so each layer can be used, checked or replaced independently.
- Pick annotation types according to which tools are reliable and which know-how is available, and favour fundamental layers over higher-level ones, as in a "minimal kit" approach.
- Extend the orthographic alphabet with a few extra marks so that every written token maps one-to-one onto a phonetic notation, taking effects across word boundaries into account.
- Have an independent party validate the resource against criteria written down in advance, covering both format and content, and report concrete deviation types and missing documentation.

## For Mova

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

- **I1** (learn): Silver parses from the witness parsers could be shown to a reviewer (or an LLM teacher) as a ranked list, with corrections focused on the minority of tokens where the top candidate is not accepted.
- **I2** (tokens): The en crate could export tokens, lemmas, morphology and tags as aligned layers, which makes each step inspectable and lets later stages be tested against a fixed earlier layer.
- **I3** (tree): When building or converting treebanks, first secure a small, very clean core layer (tokens, UPOS, basic deps) before investing in richer enhanced annotations.
- **I4** (back): A generation layer could carry explicit marks for context-dependent surface changes (e.g., a/an, contractions), so that the step from meaning graph to text stays deterministic and explainable.
- **I5** (learn): Evaluation of silver data and repair rules could use a fixed, pre-written checklist applied by a separate checker, which also produces a catalogue of recurring error types.

*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

- [University of Copenhagen](https://md.research.elamur.ai/free/en/universities/university-of-copenhagen/) — University, Copenhagen, Denmark · 115 works · 15,292 citations
- [Utrecht University](https://md.research.elamur.ai/free/en/universities/utrecht-university/) — University, Utrecht, Netherlands · 39 works · 1,603 citations
- [Sidi Mohamed Ben Abdellah University](https://md.research.elamur.ai/free/en/universities/sidi-mohamed-ben-abdellah-university/) — University, Fes, Morocco · 8 works · 156 citations
- [Evaluations and Language resources Distribution Agency](https://md.research.elamur.ai/free/en/labs/evaluations-and-language-resources-distribution-agency/) — Research institute, Paris, France · 5 works · 68 citations
- [Petra University](https://md.research.elamur.ai/free/en/universities/petra-university/) — University, Amman, Jordan · 3 works · 47 citations
- [University of Balamand](https://md.research.elamur.ai/free/en/universities/university-of-balamand/) — University, Tripoli, Lebanon · 2 works · 51 citations

📚 Shelves (4): [🖥️ 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 · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

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

## Cite

Cite as: M. Yaseen, M. Attia, B. Maegaard, K. Choukri, N. Paulsson, S. Haamid, S. Krauwer, C. Bendahman, H. Fersøe, M. Rashwan, B. Haddad, C. Mukbel, A. Mouradi, A. Al-Kufaishi, M. Shahin, N. Chenfour and A. Ragheb. 2006. Building Annotated Written and Spoken Arabic LRs in NEMLAR Project. Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06).

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

```bibtex
@inproceedings{yaseen2006building,
  title = {{Building Annotated Written and Spoken Arabic LRs in NEMLAR Project}},
  author = {M. Yaseen, M. Attia, B. Maegaard, K. Choukri, N. Paulsson, S. Haamid, S. Krauwer, C. Bendahman, H. Fersøe, M. Rashwan, B. Haddad, C. Mukbel, A. Mouradi, A. Al-Kufaishi, M. Shahin, N. Chenfour and A. Ragheb},
  booktitle = {Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06)},
  year = {2006},
  doi = {10.63317/2rxwhgo25jqb},
  url = {https://doi.org/10.63317/2rxwhgo25jqb}
}
```

- [← previous NPs for Events: Experiments in Coreference Annotation](https://md.research.elamur.ai/free/en/papers/hasler-2006-nps-events-experiments-coreference-annotation/)
- [next → Morpho-syntactic Arabic Preprocessing for Arabic to English Statistical Machine Translation](https://md.research.elamur.ai/free/en/papers/isbihani-2006-morpho-syntactic-arabic-preprocessing-arabic/)
- [↑ back to papers of 2006](https://md.research.elamur.ai/free/en/papers/2006/page/2/)

## References

14 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Attia, M. (2000). A Large-Scale Computational Processor of The Arabic Morphology, and Applications, MSc. thesis, Dept. of Computer Engineering, Faculty of Engineering, Cairo University. [http://www.RDI-eg.com/RDI/Technologies/paper.htm](http://www.RDI-eg.com/RDI/Technologies/paper.htm) [↗](http://www.RDI-eg.com/RDI/Technologies/paper.htm)
2. Attia, M., Rashwan, M., A Large-Scale Arabic POS Tagger Based on a Compact Arabic POS Tags Set, and Application on the Statistical Inference of Syntactic Diacritics of Arabic Text Words, NEMLAR int’l conference in Cairo, Sept. 2004. [http://www.RDI-eg.com/RDI/Technologies/paper.htm](http://www.RDI-eg.com/RDI/Technologies/paper.htm) [↗](http://www.RDI-eg.com/RDI/Technologies/paper.htm)
3. Attia, M. (2005), Theory and Implementation of a Large-Scale Arabic Phonetic Transcriptor, and Applications, PhD thesis, Faculty of Engineering, Dept of Electronics and Electrical Communications, Cairo University. [http://www.RDI-eg.com/RDI/Technologies/paper.htm](http://www.RDI-eg.com/RDI/Technologies/paper.htm) [↗](http://www.RDI-eg.com/RDI/Technologies/paper.htm)
4. Attia., M., Yaseen., M., Choukri., K. (2005), Specifications of the Arabic Written Corpus produced within the NEMLAR project; www.NEMLAR.org. [find](https://scholar.google.com/scholar?q=Attia%2E%2C+M%2E%2C+Yaseen%2E%2C+M%2E%2C+Choukri%2E%2C+K%2E+%282005%29%2C+Specifications+of+the+Arabic+Written+Corpus+produced+within+the+NEMLAR+project%3B+www%2ENEMLAR%2Eorg%2E)
5. Choukri., K., Paulsson., N., Haamid., S., Shahin., M. (2005), Specifications of the Arabic broadcast news speech corpus produced within the NEMLAR project, www.NEMLAR.org . [find](https://scholar.google.com/scholar?q=Choukri%2E%2C+K%2E%2C+Paulsson%2E%2C+N%2E%2C+Haamid%2E%2C+S%2E%2C+Shahin%2E%2C+M%2E+%282005%29%2C+Specifications+of+the+Arabic+broadcast+news+speech+corpus+produced+within+the+NEMLAR+project%2C+www%2ENEMLAR%2Eorg+%2E)
6. Fersøe, H. (2004). Validation Manual for Lexica. Report submitted to ELRA under the ELRA/0209/VAL-1 contract. [find](https://scholar.google.com/scholar?q=Fers%C3%B8e%2C+H%2E+%282004%29%2E+Validation+Manual+for+Lexica%2E+Report+submitted+to+ELRA+under+the+ELRA%2F0209%2FVAL%2D1+contract%2E)
7. Fersøe, H & N. Paulsson (2006). Specification of Validation Criteria. Validation Criteria for NEMLAR Arabic Written Corpus. NEMLAR report. www.NEMLAR.org . [find](https://scholar.google.com/scholar?q=Fers%C3%B8e%2C+H+%26+N%2E+Paulsson+%282006%29%2E+Specification+of+Validation+Criteria%2E+Validation+Criteria+for+NEMLAR+Arabic+Written+Corpus%2E+NEMLAR+report%2E+www%2ENEMLAR%2Eorg+%2E)
8. Haamid., S., Paulsson., N., Choukri., K., Shahin., M. (2005), Specifications of the Arabic TTS speakers DB corpus produced within the NEMLAR project, www.NEMLAR.org Krauwer, S., Maegaard, B., Choukri, K. Jørgensen, L.D. (2004), BLARK for Arabic, NEMLAR Report, University of Copenhagen. [find](https://scholar.google.com/scholar?q=Haamid%2E%2C+S%2E%2C+Paulsson%2E%2C+N%2E%2C+Choukri%2E%2C+K%2E%2C+Shahin%2E%2C+M%2E+%282005%29%2C+Specifications+of+the+Arabic+TTS+speakers+DB+corpus+produced+within+the+NEMLAR+project%2C+www%2ENEMLAR%2Eorg+Krauwer%2C+S%2E%2C+Maegaard%2C+B%2E%2C+Choukri%2C)
9. Maegaard, B., L. Damsgaard Jørgensen, S. Krauwer, K. Choukri (2004): NEMLAR: Arabic Language Resources and Tools, In: K. Choukri and B. Maegaard (ed.): Proceedings of Arabic Language Resources and Tools Conference, p. 42-54, Cairo. [find](https://scholar.google.com/scholar?q=Maegaard%2C+B%2E%2C+L%2E+Damsgaard+J%C3%B8rgensen%2C+S%2E+Krauwer%2C+K%2E+Choukri+%282004%29%3A+NEMLAR%3A+Arabic+Language+Resources+and+Tools%2C+In%3A+K%2E+Choukri+and+B%2E+Maegaard+%28ed%2E%29%3A+Proceedings+of+Arabic+Language+Resources+and+Too)
10. Maegaard, B. (2004): [NEMLAR – an Arabic Language Resources project](https://md.research.elamur.ai/free/en/papers/maegaard-2004-nemlar-arabic-language-resources-project/). In: Fourth International Conference on Language Resources and Evaluation, Proceedings Vol I, p. 109-112, Lisboa.
11. Maegaard, B, Choukri, K, Mokbel, C. and Yaseen M. (2005) Language Technology for Arabic. ISBN 87-90708-15-6, © NEMLAR, Center for Sprogteknologi, University of Copenhagen, Denmark. [find](https://scholar.google.com/scholar?q=Maegaard%2C+B%2C+Choukri%2C+K%2C+Mokbel%2C+C%2E+and+Yaseen+M%2E+%282005%29+Language+Technology+for+Arabic%2E+ISBN+87%2D90708%2D15%2D6%2C+%C2%A9+NEMLAR%2C+Center+for+Sprogteknologi%2C+University+of+Copenhagen%2C+Denmark%2E)
12. Paulsson, N. (2006-a). Specification of Validation Criteria. for NEMLAR Arabic Broadcast News Speech Corpus. Report submitted to the NEMLAR consortium. [find](https://scholar.google.com/scholar?q=Paulsson%2C+N%2E+%282006%2Da%29%2E+Specification+of+Validation+Criteria%2E+for+NEMLAR+Arabic+Broadcast+News+Speech+Corpus%2E+Report+submitted+to+the+NEMLAR+consortium%2E)
13. Paulsson, N. (2006-b). Specification of Validation Criteria for NEMLAR Arabic TTS Database. NEMLAR, www.NEMLAR.org Van den Heuvel, H, L. Boves and E. Sanders (2002). Validation of Content and Quality of SLR: Overview and Methodology. Report submitted to ELRA under the ELRA/9901/VAL-1 contract. [find](https://scholar.google.com/scholar?q=Paulsson%2C+N%2E+%282006%2Db%29%2E+Specification+of+Validation+Criteria+for+NEMLAR+Arabic+TTS+Database%2E+NEMLAR%2C+www%2ENEMLAR%2Eorg+Van+den+Heuvel%2C+H%2C+L%2E+Boves+and+E%2E+Sanders+%282002%29%2E+Validation+of+Content+and+Quality+)
14. Van den Heuvel, H. (2002) Validation criteria. Orientel. Technical Report D6.2. version 1.2 [find](https://scholar.google.com/scholar?q=Van+den+Heuvel%2C+H%2E+%282002%29+Validation+criteria%2E+Orientel%2E+Technical+Report+D6%2E2%2E+version+1%2E2)

*Updated 29 October 2026*