Understanding communication about the COVID-19 vaccines: analysis of emergent sentiments and topics of discussion on Twitter during the initial phase of the vaccine rollout

ABSTRACT

We assess the underlying topics, sentiment, and types of information regarding COVID-19 vaccines cycling through Twitter during the initiation of the vaccine rollout. Once tweets about COVID-19 vaccine posted between 1 December 2020 and 28 February 2021 were collected and preprocessed, they were categorized as either relevant or irrelevant by a classifier trained by the research team. Latent Dirichlet Allocation was used to discover the topics of discussion in the relevant tweets. The NRC lexicon was used to quantify positive and negative sentiment found in the tweets. The types of information (information, misinformation, opinion, or question) in positive and negative sentiment tweets were assessed and distributions were compared. A total of 1,386,390 tweets were collected, out of which 210,657 relevant tweets were identified by the relevancy classifier. Eight topics provided the best representation of the corpus of relevant tweets. Tweets with a negative sentiment were associated with a higher percentage of misinformation whereas tweets with positive sentiment showed a higher percentage of information, opinions, and questions. The proliferation of information and misinformation on social media platforms is associated with building public trust and mitigating negative sentiment associated with COVID-19 vaccines.

KEYWORDS:

• COVID-19 vaccine
• information
• misinformation
• sentiment analysis
• topic modeling
• Twitter

Disclosure statement

No potential conflict of interest was reported by the author(s).