pith. sign in

arxiv: 2605.20960 · v1 · pith:KOHJTPCCnew · submitted 2026-05-20 · 💻 cs.CL

JobArabi: An Arabic Corpus and Analysis of Job Announcements from Social Media

Wajdi Zaghouani , Shimaa Amer Ibrahim , Mabrouka Bessghaier , Houda Bouamor This is my paper

Pith reviewed 2026-05-21 05:42 UTC · model grok-4.3

classification 💻 cs.CL
keywords Arabic corpusjob announcementssocial mediagendered languagesociolinguistic patternsrecruitment messagesArabic NLPlabor market analysis
0
0 comments X

The pith

A new corpus of 20,528 Arabic job announcements from social media shows persistent gendered hiring language alongside regional differences in occupational demand.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper presents JobArabi, a collection of over 20,000 public job-related posts from the social media platform X gathered between January 2024 and October 2025. The data was assembled with a query system built around 21 Arabic keyword families chosen to capture gendered, plural, formal, and dialectal ways of talking about recruitment. Analysis of the posts identifies repeated patterns such as continued use of gendered terms in hiring messages, differences in job types across regions, and emotional tone in how positions are advertised. A sympathetic reader would care because the work supplies both a public dataset and concrete observations about how language in online labor markets reflects social norms and economic activity in Arabic-speaking communities.

Core claim

JobArabi is a corpus of 20,528 public posts from X collected over more than two years using a linguistically informed query framework covering 21 Arabic keyword families that reflect gendered, plural, formal, and dialectal expressions of recruitment language. Quantitative analysis of the resulting dataset, which includes posts from institutional, commercial, and individual accounts together with metadata such as timestamps and geolocation, reveals sociolinguistic patterns in online recruitment including the persistence of gendered hiring language, regional variation in occupational demand, and the emotional framing of recruitment messages. The corpus and collection scripts will be released.

What carries the argument

The JobArabi corpus assembled via a linguistically informed query framework of 21 Arabic keyword families for recruitment language.

If this is right

  • The dataset supports temporal and regional analysis of employment discourse across Arabic-speaking online communities.
  • The identified patterns demonstrate that Arabic social media can serve as a resource for studying labor market communication and linguistic change.
  • Release of the corpus enables additional work in Arabic natural language processing, computational social science, and digital labor studies.
  • Observed features such as gendered terms and emotional framing can be tracked over time within the same collection framework.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The corpus could be used to build automated detectors for biased or exclusionary language in job advertisements.
  • Regional demand differences might guide targeted training programs or migration policies in specific Arabic-speaking areas.
  • Extending collection over additional years would allow measurement of how hiring language shifts with economic or social events.
  • The same keyword-family approach could be adapted to create comparable corpora for job markets in other languages.

Load-bearing premise

The set of 21 Arabic keyword families gathers a representative sample of job announcements without major omissions or selection bias.

What would settle it

A manual review of a random sample of X posts that finds many genuine job announcements missed by the 21 keyword families, or a re-collection using broader methods that eliminates the reported gendered language patterns.

Figures

Figures reproduced from arXiv: 2605.20960 by Houda Bouamor, Mabrouka Bessghaier, Shimaa Amer Ibrahim, Wajdi Zaghouani.

Figure 4
Figure 4. Figure 4: Emotional comparison showing the distribution of Joy, Sadness, and Anger across the corpus [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
read the original abstract

This paper introduces JobArabi, a large-scale corpus of Arabic job announcements collected from social media between January 2024 and October 2025. The dataset contains 20,528 public posts from X and captures more than two years of employment-related discourse across Arabic-speaking online communities. The corpus was compiled using a linguistically informed query framework covering 21 Arabic keyword families that reflect gendered, plural, formal, and dialectal expressions of recruitment language. The resulting dataset includes posts from institutional, commercial, and individual accounts and provides metadata such as timestamps, engagement indicators, and geolocation when available, enabling temporal and regional analysis of employment discourse. Quantitative analysis reveals several sociolinguistic patterns in online recruitment, including the persistence of gendered hiring language, regional variation in occupational demand, and the emotional framing of recruitment messages. These findings highlight the potential of Arabic social media as a resource for studying labor market communication and linguistic change. The JobArabi corpus, together with documentation and collection scripts, will be released to support research in Arabic NLP, computational social science, and digital labor studies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. This paper introduces JobArabi, a corpus of 20,528 Arabic job announcements collected from X between January 2024 and October 2025 via a linguistically informed query framework of 21 Arabic keyword families (covering gendered, plural, formal, and dialectal variants). It supplies metadata including timestamps, engagement, and geolocation, and reports quantitative analysis identifying sociolinguistic patterns such as persistent gendered hiring language, regional differences in occupational demand, and emotional framing of recruitment messages, with plans to release the corpus, documentation, and scripts.

Significance. If the collection method proves representative, the work supplies a valuable large-scale resource for Arabic NLP, computational social science, and digital labor studies, enabling temporal, regional, and linguistic analyses of employment discourse that are otherwise difficult to obtain at this scale. The public release of data and scripts is a clear strength that supports reproducibility and downstream research.

major comments (2)
  1. [Corpus Construction] Corpus Construction section (description of the 21 Arabic keyword families): no recall, precision, coverage validation, manual annotation of a held-out sample, or comparison against known job-posting accounts or hashtags is reported. Without such checks, it remains unclear whether the corpus captures a representative sample of recruitment discourse or whether observed patterns (gendered language, regional demand, emotional framing) could be artifacts of keyword selection bias.
  2. [Quantitative Analysis] Quantitative Analysis section: the reported sociolinguistic patterns are presented without statistical tests, confidence intervals, effect sizes, or details on how quantitative findings (e.g., frequency counts or categorizations) were derived or validated. This absence makes it impossible to assess the robustness of the central claims about patterns in the 20,528-post corpus.
minor comments (2)
  1. [Abstract] Abstract: the collection window 'January 2024 and October 2025' spans roughly 22 months; the phrase 'more than two years' should be corrected or the exact duration clarified for precision.
  2. General: when releasing the corpus, ensure the documentation explicitly defines all metadata fields (e.g., how geolocation is determined and its coverage rate) to facilitate reuse.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback, which identifies key areas where the manuscript can be strengthened. We address each major comment below and indicate the revisions planned for the next version of the paper.

read point-by-point responses

  1. Referee: [Corpus Construction] Corpus Construction section (description of the 21 Arabic keyword families): no recall, precision, coverage validation, manual annotation of a held-out sample, or comparison against known job-posting accounts or hashtags is reported. Without such checks, it remains unclear whether the corpus captures a representative sample of recruitment discourse or whether observed patterns (gendered language, regional demand, emotional framing) could be artifacts of keyword selection bias.

    Authors: We acknowledge that the Corpus Construction section lacks explicit quantitative validation of the keyword-based collection method. The 21 Arabic keyword families were developed through iterative linguistic analysis to cover gendered, plural, formal, and dialectal variants, but we did not report precision, recall, or manual checks in the original submission. In the revised manuscript, we will add a dedicated validation subsection. This will include results from manual annotation of a random held-out sample of 500 posts by two annotators (reporting precision, inter-annotator agreement via Cohen's kappa, and examples of false positives), as well as a comparison of corpus coverage against a curated list of known recruitment accounts and high-frequency job-related hashtags on X. We will also explicitly discuss remaining limitations regarding full recall in dynamic social media environments. revision: yes

  2. Referee: [Quantitative Analysis] Quantitative Analysis section: the reported sociolinguistic patterns are presented without statistical tests, confidence intervals, effect sizes, or details on how quantitative findings (e.g., frequency counts or categorizations) were derived or validated. This absence makes it impossible to assess the robustness of the central claims about patterns in the 20,528-post corpus.

    Authors: We agree that the Quantitative Analysis section would be more robust with additional statistical detail. The original presentation focused on descriptive frequencies and qualitative interpretation of patterns such as gendered language and emotional framing. In the revised version, we will expand this section to include chi-square tests (or appropriate alternatives) for associations between variables like gender markers and regions or occupations, along with p-values, effect sizes (e.g., Cramer's V), and bootstrap-derived confidence intervals for key proportions. We will also add details on how categorizations (e.g., for emotional framing) were derived, including any manual coding procedures and validation steps. revision: yes

Circularity Check

0 steps flagged

No significant circularity in corpus construction and descriptive analysis

full rationale

The paper describes the creation of the JobArabi corpus through a linguistically informed keyword query framework and then reports quantitative descriptive statistics on sociolinguistic patterns observed in the collected posts. There are no equations, fitted parameters, predictions, uniqueness theorems, or self-citations that form a load-bearing derivation reducing to the inputs by construction. The analysis consists of direct observation and counting on the assembled dataset rather than any theoretical reduction or self-referential modeling step.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The contribution rests on the assumption that keyword-based crawling of public social media posts yields a useful sample of recruitment discourse; no free parameters, new mathematical axioms, or invented entities are introduced.

axioms (1)
  • domain assumption Public posts on X constitute accessible data for research purposes under standard platform terms.
    Implicit in the decision to collect and analyze the posts as described in the abstract.

pith-pipeline@v0.9.0 · 5732 in / 1153 out tokens · 39353 ms · 2026-05-21T05:42:01.634887+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

  1. [1]

    In Proceedings of the 2015 Conference on Em- pirical Methods in Natural Language Process - ing, pages 2515–2519, Lisbon, Portugal

    ASTD: Arabic sentiment tweets dataset . In Proceedings of the 2015 Conference on Em- pirical Methods in Natural Language Process - ing, pages 2515–2519, Lisbon, Portugal. Asso - ciation for Computational Linguistics. Elena Senger, Mike Zhang, Rob van der Goot, and Barbara Plank. 2024. Deep learning-based com- putational job market analysis: A survey on sk...

    work page 2015

  2. [2]

    arXiv preprint arXiv:2106.11040

    An exploratory study of skill require - ments for social media positions: A content analysis of job advertisements. arXiv preprint arXiv:2106.11040. Wajdi Zaghouani and Anis Charfi. 2018. Arap- tweet: A large multi -dialect twitter corpus for gender, age and language variety identifica - tion. In Proceedings of the Eleventh Interna - tional Conference on ...

    work page arXiv 2018