[Paper Review] ArCov-19: The First Arabic COVID-19 Twitter Database with Propagation Networks

Fatima Haouari, Maram Hasanain+2 more|arXiv (Cornell University)|Apr 18, 2020

Misinformation and Its Impacts21 references4 citations

TL;DR

ArCOV-19 is the first publicly available Arabic Twitter dataset covering the early COVID-19 pandemic in the Arab world, comprising ~748k popular tweets and their propagation networks (retweets and reply threads) from January 27 to March 31, 2020. It enables research in NLP, information retrieval, and social computing, with released search queries and a language-independent crawler to support future dataset curation.

ABSTRACT

In this paper, we present ArCOV-19, an Arabic COVID-19 Twitter dataset that covers the period from 27th of January till 31st of March 2020. ArCOV-19 is the first publicly-available Arabic Twitter dataset covering COVID-19 pandemic that includes around 748k popular tweets (according to Twitter search criterion) alongside the propagation networks of the most-popular subset of them. The propagation networks include both retweets and conversational threads (i.e., threads of replies). ArCOV-19 is designed to enable research under several domains including natural language processing, information retrieval, and social computing, among others. Preliminary analysis shows that ArCOV-19 captures rising discussions associated with the first reported cases of the disease as they appeared in the Arab world. In addition to the source tweets and the propagation networks, we also release the search queries and the language-independent crawler used to collect the tweets to encourage the curation of similar datasets.

Motivation & Objective

• To address the lack of publicly available Arabic social media datasets on the early stages of the COVID-19 pandemic in the Arab world.
• To provide a large-scale, high-quality dataset of Arabic tweets related to COVID-19 with structured propagation networks.
• To support research in natural language processing, information retrieval, and social computing by releasing not only the tweets but also the underlying search queries and crawler.

Proposed method

• Collection of Arabic tweets using Twitter’s search API based on predefined keywords and time constraints from January 27 to March 31, 2020.
• Application of Twitter’s popularity criterion to select the most retweeted and engaging tweets, resulting in ~748k tweets.
• Extraction of propagation networks, including both retweets and conversational reply threads, for the most popular subset of tweets.
• Design of a language-independent web crawler to facilitate future replication and curation of similar Arabic social media datasets.
• Release of the original search queries and crawler code to promote transparency and reproducibility in dataset collection.

Experimental results

Research questions

• RQ1 How did public discourse about COVID-19 evolve in the Arab world during the initial phase of the pandemic?
• RQ2 What are the structural characteristics of information propagation on Arabic Twitter during the early pandemic?
• RQ3 How do retweets and conversational threads differ in shaping the spread of COVID-19-related content in Arabic?
• RQ4 To what extent can this dataset support downstream NLP and social computing applications in low-resource languages like Arabic?
• RQ5 What methodological practices can be standardized for curating similar multilingual social media datasets?

Key findings

• The dataset captures the emergence and rise of public discussions related to the first reported cases of COVID-19 in the Arab world.
• ArCOV-19 includes approximately 748,000 popular Arabic tweets, with detailed propagation networks of retweets and reply threads.
• The dataset reveals distinct patterns in information diffusion, including rapid spread through retweets and complex conversational dynamics in replies.
• The release of search queries and a language-independent crawler enables reproducible and scalable curation of similar datasets.
• Preliminary analysis confirms that the dataset reflects real-time public sentiment and information sharing during the pandemic’s early phase.