Elamur Research⌕
☰
home / papers / 2021 / ArCOV-19: The First Arabic COVID-19 Twitter Dataset with Propagation Networks

ArCOV-19: The First Arabic COVID-19 Twitter Dataset with Propagation Networks

Paper · 2021 · Fatima Haouari, Maram Hasanain, Reem Suwaileh et al.

Summary

Arabic is one of the most-used languages on Twitter, but it was poorly covered by early COVID-19 social media collections. Some multilingual datasets held only a small Arabic share, others were random samples not focused on the pandemic, and the one Arabic-only set covered a few months and had no diffusion data. The authors built ArCOV-19, an Arabic-only COVID-19 tweet collection spanning 27 January 2020 to 31 January 2021, with about 2.7M original tweets (retweets excluded). Each day, a crawler ran hand-maintained keyword, phrase and hashtag queries against the Twitter search API. The authors revised these queries as new sub-topics trended, for example adding vaccine terms in December 2020. For a filtered set of each day's 1,000 most popular tweets, they also gathered propagation networks: retweets plus full reply threads. A first analysis showed that spikes in country-related tweets matched the first reported cases in Arab countries, and that news agencies dominated the top accounts and shared links. The resource supports NLP, IR and social computing work, especially misinformation studies.

Key points

  • Scale and coverage: about 2.7M source tweets from over 690k unique users, covering one year. Verified users make up 0.81% of accounts but wrote 18.66% of the tweets, and 25.40% of the tweets contain URLs.
  • Collection: daily Twitter search API queries were restricted to Arabic, with retweets excluded and duplicates removed within each day.
    • The API returns at most 3,200 tweets per query and only searches the past 7 days.
    • For days missed for technical reasons, the authors used GetOldTweets3 with the same queries.
  • "Top subset" selection pipeline:
    • drop tweets containing words from a custom inappropriate-word list;
    • drop tweets with more than two URLs, more than four hashtags, or fewer than four tokens;
    • for exact-text duplicates, keep only the most popular copy;
    • rank the remaining tweets by retweets plus likes and keep the top 1,000 per day.
  • Propagation networks: retweets were collected via Pickaw, and direct and indirect replies via the PHEME conversation script. At the time of writing, replies had been gathered only up to the end of April 2020.
  • Release: the dataset is distributed as tweet IDs to respect Twitter's terms. It also includes the daily query lists and a language-independent crawler. Saudi Arabia leads the geolocated content, with Kuwait second at 9.6%. The follow-up dataset ArCOV19-Rumors builds on this resource.

Interesting ideas

  • I1. Before choosing a high-value subset, the authors apply a few cheap, transparent rules (link and hashtag counts, a minimum length, a blocklist), which removes most junk without any trained classifier.
  • I2. When texts are exact duplicates, only the single highest-scoring copy is kept, so repeated content does not dominate the sample.
  • I3. The query list used for each period is published with the data, which makes the selection process auditable and reproducible even as the queries change.
  • I4. Gaps caused by a failed collector are filled by a different tool driven by the same queries, which keeps the overall sample consistent.
  • I5. Reply threads are collected as trees that include indirect replies, not just flat lists, so the conversational structure around a message is preserved.

For Mova

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

  • I1 (learn): run similar rule-based checks on candidate sentences before the witness parsers vote, so silver agreement is not spent on noisy or spam-like inputs.
  • I2 (learn): deduplicate training and evaluation sentences while keeping one representative, so frequent templates do not inflate parser accuracy or skew self-induced repair rules.
  • I3 (learn): store the exact selection criteria and source version with every silver-data batch, so a later regression can be traced to how the data was chosen.
  • I4 (learn): if a primary witness parser or teacher tier is unavailable, fall back to another one under identical inputs and filters, and mark which items came from the fallback.
  • I5 (frames): treat dialogue or multi-turn text as a reply tree, so speech-act labels and cross-turn coreference can look at the parent utterance rather than only the previous line.

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

 Institutions

🎓Qatar UniversityDoha · Qatar48 works · 1,026 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 worksCommunication · 97 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 2,908 works

Cite

Cite as Fatima Haouari, Maram Hasanain, Reem Suwaileh and Tamer Elsayed. 2021. ArCOV-19: The First Arabic COVID-19 Twitter Dataset with Propagation Networks. Proceedings of the Sixth Arabic Natural Language Processing Workshop.

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

@inproceedings{haouari2021arcov,
  title = {{ArCOV-19: The First Arabic COVID-19 Twitter Dataset with Propagation Networks}},
  author = {Fatima Haouari, Maram Hasanain, Reem Suwaileh and Tamer Elsayed},
  booktitle = {Proceedings of the Sixth Arabic Natural Language Processing Workshop},
  year = {2021},
  url = {https://aclanthology.org/2021.wanlp-1.9/}
}

References

22 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Binxuan Huang and Kathleen M Carley. 2019. A large-scale empirical study of geotagging behavior on Twitter. In Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, pages 365–373. find
  2. Sarah Alqurashi, Ahmad Alhindi, and Eisa Alanazi. 2020. Large Arabic Twitter Dataset on COVID-19. arXiv preprint arXiv:2004.04315. ↗
  3. Thayer Alshaabi, David R Dewhurst, Joshua R Minot, Michael V Arnold, Jane L Adams, Christopher M Danforth, and Peter Sheridan Dodds. 2020a. The growing echo chamber of social media: Measuring temporal and social contagion dynamics for over 150 languages on twitter for 2009–2020. arXiv preprint arXiv:2003.03667. ↗
  4. David Jurgens, Tyler Finethy, James McCorriston, Yi Tian Xu, and Derek Ruths. 2015. Geolocation prediction in twitter using social networks: A critical analysis and review of current practice. In Proceedings of the Ninth International AAAI Conference on Web and Social Media. find
  5. Ireneus Kagashe, Zhijun Yan, and Imran Suheryani. 2017. Enhancing seasonal influenza surveillance: topic analysis of widely used medicinal drugs using twitter data. Journal of medical Internet research, 19(9):e315. ↗
  6. Thayer Alshaabi, JR Minot, MV Arnold, Jane Lydia Adams, David Rushing Dewhurst, Andrew J Reagan, Roby Muhamad, Christopher M Danforth, and Peter Sheridan Dodds. 2020b. How the world’s collective attention is being paid to a pandemic: Covid-19 related 1-gram time series for 24 languages on twitter. arXiv preprint arXiv:2003.12614. ↗
  7. Christian E Lopez, Malolan Vasu, and Caleb Gallemore. 2020. Understanding the perception of covid-19 policies by mining a multilanguage twitter dataset. arXiv preprint arXiv:2003.10359. ↗
  8. Juan M Banda, Ramya Tekumalla, Guanyu Wang, Jingyuan Yu, Tuo Liu, Yuning Ding, and Gerardo Chowell. 2020. A large-scale covid-19 twitter chatter dataset for open scientific research– an international collaboration. arXiv preprint arXiv:2004.03688. ↗
  9. Andrew McNeill, Peter R. Harris, and Pam Briggs. 2016. Twitter influence on uk vaccination and antiviral uptake during the 2009 h1n1 pandemic. Frontiers in Public Health, 4:26. ↗
  10. Emily Chen, Kristina Lerman, and Emilio Ferrara. 2020a. Covid-19: The first public coronavirus twitter dataset. arXiv preprint arXiv:2003.07372. ↗
  11. Vanessa Murdock. 2011. Your mileage may vary: On the limits of social media. SIGSPATIAL Special, 3(2):62–66. find
  12. Emily Chen, Kristina Lerman, and Emilio Ferrara. 2020b. Tracking social media discourse about the covid-19 pandemic: Development of a public coronavirus twitter data set. JMIR Public Health and Surveillance, 6(2):e19273. ↗
  13. Umair Qazi, Muhammad Imran, and Ferda Ofli. 2020. Geocov19: a dataset of hundreds of millions of multilingual covid-19 tweets with location information. SIGSPATIAL Special, 12(1):6–15. find
  14. Melissa Roy, Nicolas Moreau, Cécile Rousseau, Arnaud Mercier, Andrew Wilson, and Laëtitia Atlani-Duault. 2020. Ebola and localized blame on social media: Analysis of twitter and facebook conversations during the 2014–2015 ebola epidemic. Culture, Medicine, and Psychiatry, 44(1):56–79. ↗
  15. Hassan Dashtian and Dhiraj Murthy. 2021. Cml-covid: A large-scale covid-19 twitter dataset with latent topics, sentiment and location information. arXiv preprint arXiv:2101.12202. ↗
  16. Zhiwei Gao, Shuntaro Yada, Shoko Wakamiya, and Eiji Aramaki. 2020. Naist covid: Multilingual covid-19 twitter and weibo dataset. arXiv preprint arXiv:2004.08145. ↗
  17. Lisa Singh, Shweta Bansal, Leticia Bode, Ceren Budak, Guangqing Chi, Kornraphop Kawintiranon, Colton Padden, Rebecca Vanarsdall, Emily Vraga, and Yanchen Wang. 2020. A first look at covid-19 information and misinformation sharing on twitter. arXiv preprint arXiv:2003.13907. ↗
  18. Purva Grover, Arpan Kumar Kar, Yogesh K. Dwivedi, and Marijn Janssen. 2019. Polarization and acculturation in us election 2016 outcomes – can twitter analytics predict changes in voting preferences. Technological Forecasting and Social Change, 145:438 – 460. ↗
  19. Santosh Vijaykumar, Glen Nowak, Itai Himelboim, and Yan Jin. 2018. Virtual zika transmission after the first us case: who said what and how it spread on twitter. American journal of infection control, 46(5):549–557. find
  20. Fatima Haouari, Maram Hasanain, Reem Suwaileh, and Tamer Elsayed. 2021. Arcov19-rumors: Arabic covid-19 twitter dataset for misinformation detection. In Proceedings of the Sixth Arabic Natural Language Processing Workshop, WANLP 2021. ↗
  21. Koosha Zarei, Reza Farahbakhsh, Noel Crespi, and Gareth Tyson. 2020. A first instagram dataset on covid-19. arXiv preprint arXiv:2004.12226. ↗
  22. Arkaitz Zubiaga, Maria Liakata, Rob Procter, Geraldine Wong Sak Hoi, and Peter Tolmie. 2016. Analysing how people orient to and spread rumours in social media by looking at conversational threads. PloS one, 11(3). ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/haouari-2021-arcov-19-first-arabic-covid/