COVID-19 vaccine sentiment on Twitter

From 24 January to 31 July in 2021, we collected data that anyone can view on Twitter by using the free Twitter API. By using the keywords “vaccine”, “vaccination”, “vaccinated”, “vaxxer”, “vaxxers”, “#CovidVaccine”, “covid denier”, “pfizer”, “moderna”, “astra” and “zeneca”, “sinopharm”, “sputnik”, we collected 33K tweets published by popular Twitter accounts. For each tweet, the following variables were recorded: their author (user ID), the author's categorization (healthcare professional, news media source, other accounts with thousands of followers), the date of publication (to the precision of seconds), the vaccine mentioned, the language, and the general sentiment of the tweet text on a scale from 1 to 5. For multilingual sentiment analysis, we used an open-source BERT model from Huggingface (https://huggingface.co/nlptown/bert-base-multilingual-uncased-sentiment). When multiple vaccines are mentioned in a tweet, in our data, it is recorded as multiple tweets, one for each vaccine.To uphold the privacy policy for publishing Twitter data, the tweet texts, as well as the original user identifiers for the authors of the tweets, are not disclosed. Instead, we encoded the user information with random integers. To access the complete content of these tweets, researchers may utilize the Twitter search API by referencing the provided tweet identifiers.