Behind the COVID Curtain: Analyzing Russia’s COVID-19 Response on Twitter Using Natural Language Processing and Deep Learning

Abstract

This paper analyzes the Twitter activities of five Russian political institutions, in their source languages, to assess the country’s communication response to the beginning of the COVID-19 pandemic. This study employs several Natural Language Processing techniques in English and Russian, including a deep learning model to determine tweet sentiment and a lexicon-based method to classify tweets as health related. I argue that there was a coordinated response across the different institutions. I find that most accounts in this study tweeted less during the pandemic. Further, I argue that the proportion of health-related tweets over time was guided more by political motives than health concerns. Finally, I observe a significant difference in sentiment across English and Russian tweets, with the English being more positive and the Russian being more neutral.