A large-scale analysis of COVID-19 tweets in the Arab region

Abstract

The COVID-19 virus has spread rapidly to the Arab World, affecting the public health and economy. As a result, people started communicating about the pandemic through social media such as Twitter. This paper utilizes text mining to extract useful insights into people’s perceptions and reactions to the pandemic. First, we identified 11 general topics under which COVID-19 tweets emerging from the Arab region fall. Next, we generated training data consisting of English, multidialectal Arabic, and French tweets that were manually classified into one or more of the identified 11 topics via crowdsourcing. These training data were then used to train various deep learning models to automatically classify a tweet into one or more of the 11 topics. Our best performing models were then used to perform a large-scale analysis of COVID-19 tweets emerging from the Arab region and spanning a period of over one year. Our analysis indicates that the majority of the tweets analyzed emerged from Saudi Arabia, UAE, and Egypt and that the majority of the tweets were generated by males. We also observed a surge in tweeting about all the topics as the pandemic broke followed by a slow and steady decline over the following months. We finally performed sentiment analysis on the analyzed tweets, which indicated a strong negative sentiment until mid of September 2020, after which we observed a strong positive sentiment that coincided with the surge in tweeting about vaccines.