Elamur Research⌕
☰
home / papers / 2026 / Cohesion-6K: An Arabic Dataset for Analyzing Social Cohesion and Conflict in Online Discourse

Cohesion-6K: An Arabic Dataset for Analyzing Social Cohesion and Conflict in Online Discourse

Summary

Most computational work on polarized online talk detects overt toxicity. It rarely captures the subtler balance between divisive and unifying narratives. This paper addresses that gap for Arabic with Cohesion-6K, a collection of 6,000 public Facebook posts about the Israeli Occupation of Palestine, gathered through CrowdTangle. Each post receives exactly one label on a five-step scale that moves toward cohesion, beginning at Conflict and passing through Resolution, Community Engagement and Supportive Interactions before ending at Shared Values.

Labeling mixed two routes. Expert annotators worked directly on posts. ChatGPT also proposed labels that trained humans then checked. The authors compare these manual and hybrid routes, and agreement reached Cohen's κ = 0.85.

Linking labels to engagement metadata showed that conflict posts draw roughly two to four times more interaction than resolution posts, a statistically significant gap. The authors read this as evidence that platform attention favors divisive content.

The resource matters because it gives Arabic NLP and computational social science measurable data on cohesion, beyond sentiment or hate speech. The data, guidelines and preprocessing code are planned for open release.

Key points

  • Cohesion-6K contains 6,000 Arabic public Facebook posts on the Israeli Occupation of Palestine, collected via CrowdTangle from public or verified pages only, with metadata anonymized.
  • The taxonomy has five single-label categories arranged as a continuum: Conflict, Resolution, Community Engagement, Supportive Interactions, Shared Values.
  • The annotation pipeline is hybrid: expert human labeling plus ChatGPT pre-labels verified by trained annotators. Inter-annotator agreement is Cohen's κ = 0.85.
  • Conflict-oriented posts get about 2–4× the user interaction of resolution-oriented posts (p < 0.01).
  • The planned release is under CC BY-NC-SA 4.0 and includes annotation guidelines. Limitations the authors acknowledge:
    • Facebook only;
    • a single time window;
    • one label per post;
    • possible annotator bias on irony or coded language.

Interesting ideas

  • I1. Instead of unordered classes, the labels form an ordered scale from hostile to unifying, so analyses can treat discourse function as graded rather than binary. — For Mova (frames): speech-act or stance outputs in prag could carry a position on an ordered axis (e.g., hostile→conciliatory), which makes downstream reasoning about tone explainable and comparable.
  • I2. A language model proposes labels, humans confirm or correct them, and the authors explicitly compare this hybrid route against fully manual labeling. — For Mova (learn): LLM teacher tiers could pre-label frames or trees, with a human-checked slice serving as a control that measures how much the teacher shifts the label distribution.
  • I3. Posts containing misinformation or partisan framing were kept and labeled rather than filtered out, so the corpus stays representative of the real discourse. — For Mova (learn): evaluation sets should retain messy, misleading or ironic inputs with honest labels instead of cleaning them away, so robustness is actually tested.
  • I4. The taxonomy is validated against an independent behavioral signal (engagement counts), which shows that the labels separate something measurable outside the annotation itself. — For Mova (answer): frame or world-model categories could be checked against external outcomes such as answer correctness on school tests, confirming that the distinctions matter downstream.
  • I5. The authors admit that forcing one label per post hides texts with several simultaneous functions, and they point to multi-label annotation as the remedy. — For Mova (frames): prag should allow an utterance to hold several speech acts or implicatures at once rather than picking a single winner.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Semitic languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4twhjaneai4o

@inproceedings{alathba2026cohesion,
  title = {{Cohesion-6K: An Arabic Dataset for Analyzing Social Cohesion and Conflict in Online Discourse}},
  author = {Al-Athba, Aisha Ali and Zaghouani, Wajdi},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/4twhjaneai4o},
  url = {https://doi.org/10.63317/4twhjaneai4o}
}

References

26 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mohammed Al Emadi and Wajdi Zaghouani. 2024. Emotional toll and coping strategies: Navigating the effects of annotating hate speech data. In Proceedings of the Workshop on Legal and Ethical Issues in Human Language Technologies at LREC-COLING 2024, pages 66–72. find
  2. H. Arrosvid and U. Halwati. 2021. Media framing on the palestine-israel conflict. KOMUNIKA: Jurnal Dakwah dan Komunikasi, 15(2):217–224. find
  3. Christopher A. Bail. 2021. Breaking the social media prism: How to make our platforms less polarizing. Princeton University Press. find
  4. Md Rafiqul Biswas, Fatma Mohsen, Zubair Shah, and Wajdi Zaghouani. 2023. Potentials of Chat-GPT for annotating vaccine-related tweets. In Proceedings of the 2023 Tenth International Conference on Social Networks Analysis, Management and Security (SNAMS), pages 1–6. IEEE. find
  5. Virginia Braun and Victoria Clarke. 2021. Thematic analysis: A practical guide. SAGE Publications. find
  6. Anis Charfi, Mehdi Ben-Sghaier, Amr Atalla, Ruba Akasheh, Sara Al-Emadi, and Wajdi Zaghouani. 2024. MARASTA: A multi-dialectal arabic cross-domain stance corpus. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 11060–11069. find
  7. John W. Creswell and J. David Creswell. 2018. Research design: Qualitative, quantitative, and mixed methods approaches, 5 edition. SAGE Publications. find
  8. F. Foos, L. Kostadinov, N. Marinov, and F. Schim-melfennig. 2021. Does social media promote civic activism? a field experiment with a civic campaign. Political Science Research and Methods, 9(3):500–518. find
  9. Sandra González-Bailón and Yphtach Lelkes. 2023. Do social media undermine social cohesion? a critical review. Social Issues and Policy Review, 17(1):155–180. find
  10. Ahmad Hamad Kareem and Yaseen Mahmood Najm. 2024. A critical discourse analysis of the biased role of western media in the israeli-palestinian conflict. JOURNAL OF LANGUAGE STUDIES, 8(6):200–215. find
  11. T. Khaund, B. Kirdemir, N. Agarwal, H. Liu, and F. Morstatter. 2021. Social bots and their coordination during online campaigns: A survey. IEEE Transactions on Computational Social Systems, 9(2):530–545. find
  12. Salma Laabar and Wajdi Zaghouani. 2024. Multidimensional insights: Annotated dataset of stance, sentiment, and emotion in Facebook comments on Tunisia’s July 25 measures. In Proceedings of the Second Workshop on NLP for Political Sciences at LREC-COLING 2024, pages 22–32. find
  13. J. R. Landis and G. G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33(1):159–174. find
  14. T. S. Muhammed and S. K. Mathew. 2022. The disaster of misinformation: A review of research in social media. International Journal of Data Science and Analytics, 13(4):271–285. find
  15. Kimberly A. Neuendorf. 2017. The Content Analysis Guidebook, 2 edition. SAGE Publications. find
  16. A. Raza, A. H. Hakimi, and S. Malik. 2023. Israel-palestine and social media: Sfg based critical discourse analysis of opinion article about haaretz front page. Journal of Arts and Linguistics Studies, 1(3):237–262. find
  17. S. M. A. Renhoran and H. Setiawan. 2024. Public sentiment analysis of the israel-palestine conflict on social media using bert. Indonesian Journal of Cultural and Community Development, 15(3):10–21070. find
  18. C. Russo, F. Danioni, I. Zagrean, and D. Barni. 2022. Changing personal values through value-manipulation tasks: A systematic literature review based on schwartz’s theory of basic human values. European Journal of Investigation in Health, Psychology and Education, 12(7):692– 715. find
  19. Alexey Shestakov and Wajdi Zaghouani. 2024. Analyzing conflict through data: A dataset on the digital framing of Sheikh Jarrah evictions. In Proceedings of the Second Workshop on Natural Language Processing for Political Sciences at LREC-COLING 2024, pages 55–67. find
  20. N. Stern and U. B. Shalom. 2021. Confessions and tweets: Social media and everyday experience in the israel defense forces. Armed Forces & Society, 47(2):343–366. find
  21. S. Suwarno and W. Sahayu. 2020. Palestine and israel representation in the national and international news media: A critical discourse study. Humaniora, 32(3):217–225. find
  22. Y. Tao, M. Boukes, and A. Schuck. 2024. Unpacking the nuances of agenda-setting in the online media environment: An hourly-event approach in the context of chinese economic news. Journalism Studies, 25(8):856–875. find
  23. L. T. L. Terren and R. B. B. R. Borge-Bravo. 2021. Echo chambers on social media: A systematic review of the literature. Review of Communication Research, 9. find
  24. P. Weninggalih and A. Pramiyanti. 2025. Analysis of communication patterns and social networks in digital journalism on the palestine-israel conflict. Ranah Research: Journal of Multidisciplinary Research and Development, 7(3):1594– 1602. find
  25. Wajdi Zaghouani, Hamdy Mubarak, and Md Rafiqul Biswas. 2024. So hateful! building a multi-label hate speech annotated arabic dataset. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 15044–15055. find
  26. M. Zahoor and N. Sadiq. 2021. Digital public sphere and palestine-israel conflict: A conceptual analysis of news coverage. Liberal Arts and Social Sciences International Journal (LASSI), 5(1):168–181. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/athba-2026-cohesion-6k-arabic-dataset-analyzing/