@inproceedings{zaghouani-etal-2026-nakba,
    title = "Nakba Discourse 2025: A Bilingual Social Media Dataset for Collective Trauma Analysis",
    author = "Zaghouani, Wajdi  and
      Bessghaier, Mabrouka  and
      Attia, Kais",
    editor = "Jarrar, Mustafa  and
      El-Haj, Mo  and
      Haddad, Amal  and
      Atiani, Serin  and
      Abudalfa, Shadi  and
      Regier, Terry  and
      Rayson, Paul  and
      Sima{'}an, Khalil  and
      Mansour, Camille",
    booktitle = "Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ {LREC} 2026",
    month = may,
    year = "2026",
    address = "Palma, Mallorca (Spain)",
    publisher = "ELRA Language Resources Association (ELRA)",
    url = "https://preview.aclanthology.org/cawl-year/2026.nakbanlp-1.3/",
    doi = "10.63317/26i8wdd5eyrt",
    pages = "33--42",
    abstract = "We introduce Nakba Discourse 2025, a bilingual full-year social media dataset capturing Arabic and English discourse about the 1948 Palestinian Nakba across Twitter/X and Facebook from January to December 2025. The corpus contains 70,312 unique posts organized into intersecting sub-corpora by language, sentiment, gender, geography, and platform, with engagement metadata and automatically extracted rhetorical features. Analyses reveal systematic variation in engagement and framing across communities. Per-post engagement is highest in Israel and UK subsets (50.62 and 49.08 average likes respectively), while Arabic-language discourse shows markedly lower per-post engagement. Sentiment distribution is strongly skewed, with negative sentiment posts outnumbering positive ones at an 11:1 ratio (54,424 vs. 4,827 posts). Despite dramatic variation in absolute engagement levels, virality rates remain structurally constant at approximately 10{\%} across all Twitter/X sub-corpora, regardless of language, gender, or geography, pointing to platform-level amplification regularities. Gender analysis reveals that women achieve proportional virality equal to men despite producing roughly one-third the volume of posts. Temporal patterns align with cultural calendars, including Thursday peaks associated with Jumu{'}ah across Arabic and female subsets, and Sunday peaks in English-language subsets reflecting Western media cycles. The dataset will be released for research use and supports multilingual stance detection, virality modeling, rhetorical analysis, and computational studies of digital political memory."
}