language: ar
license: cc-by-4.0
tags:
- arabic
- jobs
- recruitment
- gender
- sociolinguistics
- labour-market
task_categories:
- text-classification
- summarization
- question-answering
- token-classification
size_categories:
- 5K<n<10K
pretty_name: 'ArabJobs: A Multinational Corpus of Arabic Job Advertisements'
dataset_info:
features:
- name: job_title
dtype: string
description: Job title as written in the advertisement.
- name: location
dtype: string
description: City or region where the job is located.
- name: salary
dtype: string
description: Raw salary text exactly as written in the job post.
- name: profession
dtype: string
description: Derived profession label inferred from job description.
- name: description
dtype: string
description: Full job description text.
- name: gender
dtype: string
description: Gender signals extracted from linguistic markers (male/female/neutral).
- name: country
dtype: string
description: >-
Country from which the job post originates (Egypt, Jordan, Saudi Arabia,
UAE).
- name: salary_local
dtype: float32
description: Salary normalised into local currency.
- name: salary_usd
dtype: float32
description: Salary converted into USD.
- name: job_category
dtype: string
description: High-level job category (e.g. IT, Finance, Education).
- name: sub_category
dtype: string
description: More specific category inside the main job category.
splits:
- name: train
- name: validation
- name: test
ArabJobs: A Multinational Corpus of Arabic Job Advertisements
๐ Overview
ArabJobs is the first publicly available, multinational corpus of Arabic job advertisements, collected from
Egypt, Jordan, Saudi Arabia, and the UAE.
It contains:
- 8,546 job postings
- 550,000+ words
- Coverage across numerous sectors and dialects
- Rich metadata including salary, profession, gender indicators, and job categories
This dataset supports research on:
- Fairness-aware Arabic NLP
- Gender representation and bias in recruitment language
- Dialectal variation and code-switching
- Salary estimation and profession classification
- Labour market discourse across Arab countries
๐ Citation
If you use ArabJobs, please cite:
El-Haj, M. (2025). ArabJobs: A Multinational Corpus of Arabic Job Ads.
Proceedings of the 3rd ArabicNLP Conference, EMNLP, Suzhou, China, 9โ10 November 2025.
BibTeX
@inproceedings{elhaj2025arabjobs,
title={ArabJobs: A Multinational Corpus of Arabic Job Ads},
author={El-Haj, Mo},
booktitle={Proceedings of the ArabicNLP Workshop, EMNLP},
year={2025},
address={Suzhou, China},
month={November}
}
๐ Dataset Structure
Each row in the dataset corresponds to a single job advertisement with the following key fields:
job_titleโ the job title in Arabicdescriptionโ full job description textcountryโ country of postinglocationโ city or regionsalaryโ raw salary textsalary_local,salary_usdโ normalised salaryprofessionโ derived profession labelgenderโ male, female, or neutral (when detectable)job_category,sub_categoryโ hierarchical job categories
All text is encoded in UTF-8 Arabic.
๐ Visualisations
Gendered Job Advertisements (Word Clouds)
Job Ads by Country
Dialectal Variation
๐งช Example Records
Here are sample entries from the dataset:
| job_title | location | salary | profession | description | gender | country | salary_local | salary_usd | job_category | sub_category |
|---|---|---|---|---|---|---|---|---|---|---|
| ู ุฏูุฑ ู ูุชุจ ููู (ุงุซุงุซ ) (ูููู ุชุทุจูููู ) - 15 ู ุงูู - ุงููุงูุฑุฉ | 15 ู ุงูู - ุงููุงูุฑุฉ | 15000 - 25000 ุฌููู ู ุตุฑู | ููุฏุณุฉ ุนู ุงุฑุฉ ูุฏูููุฑ | ู ุทููุจ ู ุฏูุฑ ู ูุชุจ ููู ุฎุจุฑุฉ ูุง ุชูู ุนู 7 ุณููุงุชโฆ | neutral | Egypt | 20000 | 640 | ุฅุนูุงู ูุชุตู ูู | ุชุตู ูู |
| ูุธููุฉ ุฃุฎุตุงุฆู ู ูุงูุน ุฅููุชุฑูููุฉ | ุงูุฑูุงุถ | 4,000/ุดูุฑูุงู | ุฃุฎุตุงุฆู ู ูุงูุน ุฅููุชุฑูููุฉ | ูุนูู ุตูุฏูู ุงุณุฑุฉ ุงูุฎุถูุฑ ุนู ุญุงุฌุชู ูู ูุธู/ุฃู ู ูุธูุฉโฆ | female | Saudi Arabia | 4000 | 1080 | ุชูููููุฌูุง ุงูู ุนููู ุงุช | ุชูููููุฌูุง ุงูู ุนููู ุงุช |
| ู ุนูู ุนุฑุจู ุดุงูุฑู ุง ุนุฏุฏ 2 | ุงูุฒุฑูุงุก ุงูุฌุฏูุฏุฉ | โ | ููุฏูุฉ ูู ุทุงุนู ุ ู ุนูู ุดุงูุฑู ุง | ู ุทููุจ ูุฏู ู ุทุงุนู ุงุจู ูููุณ ุงูุฒุฑูุงุก ุงูุฌุฏูุฏุฉโฆ | male | Jordan | 420 | 592.2 | ุณูุงุญุฉ ูู ุทุงุนู | ููุฏูุฉ ูู ุทุงุนู |
| ูุธุงุฆู ุญุณุงุจุงุช ูู ุฅู ุงุฑุฉ ุงูุดุงุฑูุฉ ุงูุฅู ุงุฑุงุช | Sharjah | 3000 ุฏุฑูู | ุญุณุงุจุงุช | ูุธููุฉ ุดุงุบุฑุฉ - ู ูุธูุฉ ุญุณุงุจุงุชโฆ | neutral | UAE | 3000 | 810 | ู ุงููุฉ ูู ุญุงุณุจุฉ | ู ุงููุฉ ูู ุญุงุณุจุฉ |
๐ License
This dataset is released under the CC BY 4.0 License, allowing research and academic use with appropriate attribution.
๐ค Contributions
Feedback and contributions are welcome.
Please open an issue or pull request on the GitHub repository: