Tweeting on COVID-19 pandemic in South Africa: LDA-based topic modelling approach

Abstract

The advent of COVID-19 has disrupted all facets of human lives. As of September 2020, there is no effective viral therapy for the disease, thus necessitating research efforts toward providing solutions to the diverse areas where the pandemic has wreaked havoc. As a way of reducing the spread of the disease, the South African government declared COVID-19 a national disaster and implemented nationwide lockdowns with several regulations. Nevertheless, the success of such synergized efforts primarily depends on the people’s attitudes and perceptions toward the multifaceted management of the pandemic. Therefore, this current study aims to discover what topical issues relating to the pandemic are being discussed by the populace and what impacts these issues have on compliance with regulations, including how they can aid in the implementation of the measures put in place by the government, as we analyze discussions relating to COVID-19 using data harvested from Twitter – social media and opinion mining platform. The Latent Dirichlet Allocation (LDA) algorithm was applied for the extraction of noteworthy topics. From the experiments conducted, it was observed that alcohol sale and consumption, staying home, daily statistics tracing, police brutality, 5G and vaccines conspiracy theories were among the topics discussed and around which attitudes and perceptions were formed by the citizens. The findings also revealed people’s resistance to measures that affect their economic activities, and their unwillingness to take tests or vaccines as a result of fake news and conspiracy theories. These findings can assist the government and policymakers in redirecting their efforts by addressing the citizens’ concerns and reactions to the instituted measures toward an anticipated overall success.