Abstract

Purpose This study aims to understand the relationship between politics and pandemics in shaping the characteristics and themes of people’s Tweets during the US 2020 presidential election. Additionally, the purpose is to detect misinformation and extremism, not only to help online social networks to target such content more rapidly but also to provide a close to real-time picture of trending topics, misinformation, and extremism flowing on OSN. This could help authorities to identify the intents behind them and find out how and when they should address such content. Design/methodology/approach This study focuses on extracting and verifying knowledge from large-scale OSN data, at the intersection of the Coronavirus pandemic and the US 2020 presidential election. More specifically, this study makes manual, statistical and automatic inferences and extracts knowledge from over a million Tweets related to the two aforementioned major events. On the other hand, disinformation operations intensified in 2020 with the coincidence of the Coronavirus pandemic and presidential election. This study applies machine learning to detect misinformation and extreme opinions on OSN. Over one million Tweets have been collected by our server in real-time from the beginning of April 2020 to the end of January 2021, using six keywords, namely, Covid, Corona, Trump, Biden, Democrats and Republicans. These Tweets are inspected with regard to their topics, opinions, news, and political affiliation, along with misinformation and extremism. Findings Our analyses showed that the majority of these Tweets concern death tolls, testing, mask, drugs, vaccine, and travel bans. The second concern among these Tweets is reopening the economy and schools, unemployment, and stimulus bills. The third concern is related to the Coronavirus pandemic’s impacts on politics, voting, and misinformation. This highlights the topics that US voters on Twitter were most concerned about during this time period, among the multitude of other topics that politicians and news media were reporting or discussing. Automatic classification of these Tweets using a long short-term memory network revealed that Tweets containing misinformation formed between 0.5% and 1.1% of Coronavirus-related Tweets every month and Tweets containing extreme opinions formed between 0.5% and 3.1% of them every month, with its pick in October 2020, coinciding with the US presidential election month. Originality/value The originality of this study lies in establishing a framework to collect, process, and classify OSN data to detect misinformation and extremism and to provide a close to real-time picture of trending topics, misinformation, and extremism flowing on OSN.