A Text Mining Analysis on Misinformation Regarding the COVID-19 Pandemic

Abstract

Since the outset of COVID-19 pandemic, a massive amount of information has been generated about the pandemic, where a great deal of it contains less verifiable information disseminated especially via social media. A video propagating various conspiracy theories about the pandemic, called plandemic, was launched, and people started to share posts addressing this issue with this hashtag thereafter. For this research, we collected thousands of tweets using this hashtag, and then combined this collection with a collection of tweets with a similar hashtag #scamdemic to build a study group. Also, we collected tweets that convey more general thoughts about the pandemic, which served as a control group. We showed that the web sources provided in the tweets in the study group tend to be much less credible. Furthermore, we performed two sentiment analysis using Hedonometer and VADER. Hedonometer showed that the average happiness level in tweets spreading misinformation about COVID -19 is almost the same as in regular COVID -19 tweets. However, VADER showed that the tweets spreading the misinformation have significantly more negative sentiment. This could be related to the fact that the VADER also takes into account non-lexical items, such as emoticons and capital letters.