---
title: "ArabJobs: A Multinational Corpus of Arabic Job Ads (2025)"
description: "Arabic NLP has no open collection of job advertisements, although English recruitment texts have… 11 references, cited by 5 papers here."
authors: ["Mo El-Haj"]
year: 2025
venue: "Proceedings of The Third Arabic Natural Language Processing Conference"
source: "https://aclanthology.org/2025.arabicnlp-main.2/"
doi: "10.18653/v1/2025.arabicnlp-main.2"
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/haj-2025-arabjobs-multinational-corpus-arabic-job/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/haj-2025-arabjobs-multinational-corpus-arabic-job/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/page/11/) / ArabJobs: A Multinational Corpus of Arabic Job Ads

# ArabJobs: A Multinational Corpus of Arabic Job Ads

**Authors:** [Mo El-Haj](https://md.research.elamur.ai/free/en/authors/mo-el-haj/) · **Year:** 2025 · **Venue:** [Proceedings of The Third Arabic Natural Language Processing Conference](https://md.research.elamur.ai/free/en/venues/arabicnlp/) ([ArabicNLP 2025](https://md.research.elamur.ai/free/en/venues/arabicnlp/2025/)) · **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/2025.arabicnlp-main.2/) · [PDF](https://aclanthology.org/2025.arabicnlp-main.2.pdf) · [DOI 10.18653/v1/2025.arabicnlp-main.2](https://doi.org/10.18653/v1/2025.arabicnlp-main.2) · [OpenAlex](https://openalex.org/W4416037682)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · [5](https://md.research.elamur.ai/free/en/papers/haj-2025-arabjobs-multinational-corpus-arabic-job/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Arabic NLP has no open collection of job advertisements, although English recruitment texts have been widely used to study bias and labour-market discourse. This paper fills that gap with ArabJobs, a public corpus of 8,546 Arabic job ads containing over 550,000 words. The ads were scraped, with robots.txt and rate limits respected, from seven recruitment sites covering Egypt, Jordan, Saudi Arabia and the UAE. Each record has scraped fields such as title, location, salary, gender target and description. It also has fields inferred by an LLM and then checked by native speakers.

The author explores regional dialect differences and English code-switching. GPT-4 is used with few-shot prompting to fill in missing salaries and to normalise job categories. The paper also measures gender skew in ads and pay, and lists the kinds of discriminatory hiring criteria the ads contain. GPT-4's salary estimates match known values closely and agree with human annotators at levels close to human–human agreement. The corpus supports fairness-oriented Arabic NLP and labour research in a domain grounded in real-world use.

## Key points
- The corpus has 8,546 ads: Egypt 3,598, UAE 1,947, Saudi Arabia 1,854 and Jordan 1,147. Average ad length ranges from about 29 words (UAE) to about 117 words (Saudi Arabia). Male-targeted ads outnumber female-targeted ones in every country.
- The sources are naukrigulf, gulftalent, dubizzle, tanqeeb, jordanrec, forasna and sabbar. Personal identifiers were removed, and the corpus is released on GitHub under a non-commercial, research-only licence.
- Only 3,265 ads state a salary. On those ads, GPT-4 with 3 in-context examples reached MAE 11.83, RMSE 14.84 and Pearson 0.997, with 98.5% of estimates within ±10% of the true value.
- Two annotators each labelled 500 ads. They agreed with each other 93% of the time within ±20% and 89% within ±10%. GPT-4 agreed with them 85%/81% (Annotator 1) and 81%/78% (Annotator 2) at the same margins.
- Dialects were explored by sampling 1,500 ads per country and projecting TF-IDF vectors with Truncated SVD. Saudi and Emirati ads form one cluster, while Egyptian and Jordanian ads sit apart. Jordanian, Egyptian and Saudi ads mix in more English job terms than UAE ads.

## Interesting ideas
- The LLM-inferred salaries are evaluated two ways: against the ads with a stated salary, and against humans, using tolerance bands and the human–human agreement as a ceiling. This shows whether the model's error is close to the inherent ambiguity of the task.
- Model-derived columns (profession, salary_local, salary_usd, job_category, sub_category) are kept separate from scraped columns and then checked by native speakers. This keeps track of which values were stated in the source and which were guessed.
- For pay comparisons, the author computes male and female averages per category and gives more weight to the dominant gender. This avoids a few unusual ads from the minority group distorting the picture in heavily skewed categories.
- Discriminatory criteria fall into a compact typology covering physical looks, character traits, age restrictions, requirements about marital status or availability, filtering by gender, and framing around emotions or soft skills. Each type is backed by concordance examples from the ads.
- A cheap unsupervised projection (TF-IDF plus SVD over balanced per-region samples) is used as a sanity check that regional variation really exists before any annotation is done.

## For Mova

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

- **I1** (learn): when grading LLM teacher tiers or silver labels, report agreement with humans next to human–human agreement, so a teacher is judged against a realistic ceiling.
- **I2** (world): tag each entity attribute in the text-world as stated or inferred, with its source, so answers can explain whether a value came from the text or was filled in.
- **I3** (learn): for per-label or per-construction scores on imbalanced data, report sub-group figures with their support counts and weight by dominance, so rare cases don't mislead error analysis.
- **I4** (frames): treat requirement statements as constraint frames placed on a candidate role, with typed slots such as age, gender and appearance. Such requirements could then be detected and explained, rather than just spotted as keywords.
- **I5** (tree): before mixing treebanks or converting between UD dialects, cluster balanced samples to see which sources differ in genre or style, which helps decide how to split training and evaluation data.

*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 (6): [🖥️ 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 · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,772 works · [🤝 Safety Research](https://md.research.elamur.ai/free/en/shelves/social-sciences/safety-research/) · 543 works

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

## Cite

Cite as: Mo El-Haj. 2025. ArabJobs: A Multinational Corpus of Arabic Job Ads. Proceedings of The Third Arabic Natural Language Processing Conference.

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

```bibtex
@inproceedings{elhaj2025arabjobs,
  title = {{ArabJobs: A Multinational Corpus of Arabic Job Ads}},
  author = {Mo El-Haj},
  booktitle = {Proceedings of The Third Arabic Natural Language Processing Conference},
  year = {2025},
  doi = {10.18653/v1/2025.arabicnlp-main.2},
  url = {https://doi.org/10.18653/v1/2025.arabicnlp-main.2}
}
```

- [← previous Analysing Chain of Thought Dynamics: Active Guidance or Unfaithful Post-hoc Rationalisation?](https://md.research.elamur.ai/free/en/papers/lim-2025-analysing-chain-thought-dynamics-active/)
- [next → Beyond Gold Standards: Epistemic Ensemble of LLM Judges for Formal Mathematical Reasoning](https://md.research.elamur.ai/free/en/papers/zhang-2025-gold-standards-epistemic-ensemble-llm/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/page/11/)

## References

11 references, 11 of them with a card of ours. ↗ leads to the work.

1. Muhammad Abdul-Mageed, Chiyu Zhang, Houda Bouamor, and Nizar Habash. 2020. [NADI 2020: The first nuanced Arabic dialect identification shared task](https://md.research.elamur.ai/free/en/papers/mageed-2020-nadi-2020-first-nuanced-arabic/). In Proceedings of the Fifth Arabic Natural Language Processing Workshop, pages 97–110, Barcelona, Spain (Online). Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.2010.11334)
2. Khalid N Elmadani, Nizar Habash, and Hanada Taha-Thomure. 2025. [A large and balanced corpus for fine-grained arabic readability assessment](https://md.research.elamur.ai/free/en/papers/elmadani-2025-large-balanced-corpus-fine-grained/). arXiv preprint arXiv:2502.13520. [↗](https://arxiv.org/abs/2502.13520)
3. Bashar Alhafni, Nizar Habash, and Houda Bouamor. 2022. [The Arabic parallel gender corpus 2.0: Extensions and analyses](https://md.research.elamur.ai/free/en/papers/alhafni-2022-arabic-parallel-gender-corpus-2/). In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 1870–1884, Marseille, France. European Language Resources Association.
4. Amr Keleg, Sharon Goldwater, and Walid Magdy. 2023. [ALDi: Quantifying the Arabic level of dialectness of text](https://md.research.elamur.ai/free/en/papers/keleg-2023-aldi-quantifying-arabic-level-dialectness/). In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 10597–10611, Singapore. Association for Computational Linguistics.
5. Bashar Alhafni, Reem Hazim, Juan Piñeros Liberato, Muhamed Al Khalil, and Nizar Habash. 2024. [The samer arabic text simplification corpus](https://md.research.elamur.ai/free/en/papers/alhafni-2024-samer-arabic-text-simplification-corpus/). arXiv preprint arXiv:2404.18615. [↗](https://arxiv.org/abs/2404.18615)
6. Anne Lauscher, Rafik Takieddin, Simone Paolo Ponzetto, and Goran Glavaš. 2020. [AraWEAT: Multidimensional analysis of biases in Arabic word embeddings](https://md.research.elamur.ai/free/en/papers/lauscher-2020-araweat-multidimensional-analysis-biases-arabic/). In Proceedings of the Fifth Arabic Natural Language Processing Workshop, pages 192–199, Barcelona, Spain (Online). Association for Computational Linguistics.
7. Wissam Antoun, Fady Baly, and Hazem Hajj. 2020. [Arabert: Transformer-based model for arabic language understanding](https://md.research.elamur.ai/free/en/papers/antoun-2020-arabert-transformer-based-model-arabic/). In Proceedings of the 4th Workshop on Open-Source Arabic Corpora and Processing Tools, pages 9–15. [↗](https://doi.org/10.48550/arxiv.2003.00104)
8. Houda Bouamor, Sabit Hassan, and Nizar Habash. 2019. [The MADAR shared task on Arabic fine-grained dialect identification](https://md.research.elamur.ai/free/en/papers/bouamor-2019-madar-shared-task-arabic-fine/). In Proceedings of the Fourth Arabic Natural Language Processing Workshop, pages 199–207, Florence, Italy. Association for Computational Linguistics.
9. Ossama Obeid, Nasser Zalmout, Salam Khalifa, Dima Taji, Mai Oudah, Bashar Alhafni, Go Inoue, Fadhl Eryani, Alexander Erdmann, and Nizar Habash. 2020. [Camel tools: An open source python toolkit for arabic natural language processing](https://md.research.elamur.ai/free/en/papers/obeid-2020-camel-tools-open-source-python/). In Proceedings of the twelfth language resources and evaluation conference, pages 7022–7032.
10. Mouath Abu Daoud, Chaimae Abouzahir, Leen Kharouf, Walid Al-Eisawi, Nizar Habash, and Farah E Shamout. 2025. [Medarabiq: Benchmarking large language models on arabic medical tasks](https://md.research.elamur.ai/free/en/papers/daoud-2025-medarabiq-benchmarking-large-language-models/). arXiv preprint arXiv:2505.03427. [↗](https://arxiv.org/abs/2505.03427)
11. Malika Dikshit, Houda Bouamor, and Nizar Habash. 2024a. Investigating gender bias in stem job advertisements. In Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing (GeBNLP), pages 1–10. [Gpt-4 technical report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/).

*Updated 1 November 2026*