---
title: "Sudanese-Flores: Extending FLORES+ to Sudanese Arabic Dialect"
description: "Sudanese Arabic has more than 40 million speakers, yet it had no public machine translation… 10.4 BLEU. 7 references, cited by 1 paper here."
authors: ["Hadia Mohmmedosman Ahmed Samil", "David Ifeoluwa Adelani"]
year: 2026
venue: "Proceedings of the 7th Workshop on African Natural Language Processing (AfricaNLP 2026)"
source: "https://aclanthology.org/2026.africanlp-main.25/"
doi: "10.18653/v1/2026.africanlp-main.25"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/samil-2026-sudanese-flores-extending-flores-sudanese/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/samil-2026-sudanese-flores-extending-flores-sudanese/)

[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/17/) / Sudanese-Flores: Extending FLORES+ to Sudanese Arabic Dialect

# Sudanese-Flores: Extending FLORES+ to Sudanese Arabic Dialect

**Authors:** Hadia Mohmmedosman Ahmed Samil and [David Ifeoluwa Adelani](https://md.research.elamur.ai/free/en/authors/david-ifeoluwa-adelani/) · **Year:** 2026 · **Venue:** [Proceedings of the 7th Workshop on African Natural Language Processing (AfricaNLP 2026)](https://md.research.elamur.ai/free/en/venues/africanlp/) ([Workshop on African Natural Language Processing 2026](https://md.research.elamur.ai/free/en/venues/africanlp/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.africanlp-main.25/) · [PDF](https://aclanthology.org/2026.africanlp-main.25.pdf) · [DOI 10.18653/v1/2026.africanlp-main.25](https://doi.org/10.18653/v1/2026.africanlp-main.25) · [OpenAlex](https://openalex.org/W7140116280)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Sudanese Arabic has more than 40 million speakers, yet it had no public machine translation benchmark. Large multilingual resources such as Flores-200, SIB-200, FineWeb2 and MADLAD-400 leave it out, and the existing dialect corpora are small and built for morphology rather than MT.

To fill this gap, the authors created Sudanese-Flores. A trilingual native speaker rendered the Modern Standard Arabic DEV and DEVTEST portions of Flores+ into the Khartoum/Central Sudanese variety. She checked the English originals whenever the MSA was unclear, and a second native speaker audited a random sample. Because the set is aligned with Flores+, it connects Sudanese Arabic to over 200 other languages.

The authors tested GPT-4.1, Gemini 2.5 Flash, Llama-3.3 70B and Qwen-3-Next-80B-A3B with one shared, simple prompt in zero-shot and 5-shot modes, scoring outputs with BLEU and ChrF++. Models understand the dialect fairly well but produce it poorly. Generating Sudanese from English trails generating MSA by a wide margin, and few-shot examples bring small or inconsistent gains. The benchmark therefore exposes a concrete weakness of current LLMs on dialect generation.

## Key points
- **Dataset size and composition:** 2,009 sentences in total, 1,012 in DEV and 997 in DEVTEST.
  - Translation took 3–4 weeks.
  - 100 randomly chosen sentences were verified by a second native speaker.
  - The authors note that Sudanese spelling is not standardized.
- **Evaluation setup:** six directions among arb (MSA), apd (Sudanese Arabic) and eng, reported on DEVTEST. The prompt template is taken from AfroBench.
- **English → Sudanese is the hardest direction (zero-shot):**
  - GPT-4.1: 10.4 BLEU into Sudanese vs. 27.9 into MSA.
  - Gemini 2.5 Flash: 9.7 vs. 23.5.
- **Sudanese → English is much easier:** GPT-4.1 reaches 32.3 BLEU.
- **Model comparisons:**
  - Proprietary models generally outperform the open-weight ones.
  - With 5 shots, Llama-3.3 70B improves markedly on arb→apd, from 17.7 to 29.3 BLEU.
  - Qwen's few-shot gains are small and irregular.
- **Error analysis:** the most common problems are mostly stylistic rather than semantic. They include:
  - MSA forms creeping into outputs translated from Sudanese;
  - inconsistent dialect word choices;
  - weakened Sudanese idioms;
  - tense and punctuation slips.

*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 (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: Hadia Mohmmedosman Ahmed Samil and David Ifeoluwa Adelani. 2026. Sudanese-Flores: Extending FLORES+ to Sudanese Arabic Dialect. Proceedings of the 7th Workshop on African Natural Language Processing (AfricaNLP 2026).

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

```bibtex
@inproceedings{samil2026sudanese,
  title = {{Sudanese-Flores: Extending FLORES+ to Sudanese Arabic Dialect}},
  author = {Hadia Mohmmedosman Ahmed Samil and David Ifeoluwa Adelani},
  booktitle = {Proceedings of the 7th Workshop on African Natural Language Processing (AfricaNLP 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.africanlp-main.25},
  url = {https://doi.org/10.18653/v1/2026.africanlp-main.25}
}
```

- [← previous Subevent Structure as a Predictor of Entity Identity Change in Procedural Text](https://md.research.elamur.ai/free/en/papers/rim-2026-subevent-structure-as-predictor-entity/)
- [next → Supercharging Agenda Setting Research: The ParlaCAP Dataset of 28 European Parliaments and a Scalable Multilingual LLM-Based Classification](https://md.research.elamur.ai/free/en/papers/pungersek-2026-supercharging-agenda-setting-research-parlacap/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/17/)

## References

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

1. David Ifeoluwa Adelani, Hannah Liu, Xiaoyu Shen, Nikita Vassilyev, Jesujoba O. Alabi, Yanke Mao, Haonan Gao, and En-Shiun Annie Lee. 2024. [SIB-200: A simple, inclusive, and big evaluation dataset for topic classification in 200+ languages and dialects](https://md.research.elamur.ai/free/en/papers/adelani-2023-sib-200-simple-inclusive-big/). In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 226–245, St. Julian’s, Malta. Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.2309.07445)
2. Mustafa Jarrar, Fadi A. Zaraket, Tymaa Hammouda, Daanish Masood Alavi, and Martin Wählisch. 2022. [Lîsan: Yemeni, iraqi, libyan, and sudanese arabic dialect corpora with morphological annotations](https://md.research.elamur.ai/free/en/papers/jarrar-2022-lisan-yemeni-iraqi-libyan-sudanese/). Preprint, arXiv:2212.06468. Accessed: 2025-12-20. [↗](https://arxiv.org/abs/2212.06468)
3. Sneha Kudugunta, Isaac Rayburn Caswell, Biao Zhang, Xavier Garcia, Derrick Xin, Aditya Kusupati, Romi Stella, Ankur Bapna, and Orhan Firat. 2023. [MADLAD-400: A multilingual and document-level large audited dataset](https://md.research.elamur.ai/free/en/papers/kudugunta-2023-madlad-400-multilingual-document-level/). In Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track.
4. NLLB-Team, Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Guillaume Wenzek, Al Youngblood, Bapi Akula, Loic Barrault, Gabriel Mejia Gonzalez, Prangthip Hansanti, and 20 others. 2022. No language left behind: Scaling human-centered machine translation. [find](https://scholar.google.com/scholar?q=NLLB%2DTeam%2C+Marta+R%2E+Costa%2Djuss%C3%A0%2C+James+Cross%2C+Onur+%C3%87elebi%2C+Maha+Elbayad%2C+Kenneth+Heafield%2C+Kevin+Heffernan%2C+Elahe+Kalbassi%2C+Janice+Lam%2C+Daniel+Licht%2C+Jean+Maillard%2C+Anna+Sun%2C+Skyler+Wang%2C+Guillaume+We)
5. Jessica Ojo, Odunayo Ogundepo, Akintunde Oladipo, Kelechi Ogueji, Jimmy Lin, Pontus Stenetorp, and David Ifeoluwa Adelani. 2025. AfroBench: How good are large language models on African languages? In Findings of the Association for Computational Linguistics: ACL 2025, pages 19048–19095, Vienna, Austria. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Jessica+Ojo%2C+Odunayo+Ogundepo%2C+Akintunde+Oladipo%2C+Kelechi+Ogueji%2C+Jimmy+Lin%2C+Pontus+Stenetorp%2C+and+David+Ifeoluwa+Adelani%2E+2025%2E+AfroBench%3A+How+good+are+large+language+models+on+African+languages%3F+In+)
6. Guilherme Penedo, Hynek Kydlíček, Vinko Sabolčec, Bettina Messmer, Negar Foroutan, Amir Hossein Kargaran, Colin Raffel, Martin Jaggi, Leandro Von Werra, and Thomas Wolf. 2025. [Fineweb2: One pipeline to scale them all — adapting pre-training data processing to every language](https://md.research.elamur.ai/free/en/papers/penedo-2025-fineweb2-one-pipeline-scale-them/). In Second Conference on Language Modeling. [↗](https://doi.org/10.48550/arxiv.2506.20920)
7. Maja Popović. 2017. chrF++: Towards character and word n-gram F-score for automatic mt evaluation. In Proceedings of the Second Conference on Machine Translation, pages 568–575. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Maja+Popovic%CC%81%2E+2017%2E+chrF%2B%2B%3A+Towards+character+and+word+n%2Dgram+F%2Dscore+for+automatic+mt+evaluation%2E+In+Proceedings+of+the+Second+Conference+on+Machine+Translation%2C+pages+568%E2%80%93575%2E+Association+for+Compu)

*Updated 1 November 2026*