Fang, Z., & Costas, R. (2020). Tracking the Twitter attention around the research efforts on the COVID-19 pandemic. arXiv preprint arXiv:2006.05783. https://arxiv.org/abs/2006.05783

University of California, Berkeley

Google

This study looked at the Twitter attention on COVID-19 research. They saw that the number of Twitter mentions of COVID-19 related publications is increasing, with influences from scholarly-oriented Twitter users who disseminate research information. They also found that the focus of discussion has changed from more virological and clinical research to potential treatments, government action, healthcare measures, and the economy. 
They looked at the following four research questions: 
- "How are the COVID-19-related scientific publications mentioned on Twitter over time since the outbreak?"
- "By drawing on the Twitter users’ profile descriptions, who are the actors tweeting COVID-19-related research? What kinds of Twitter users are more influential in terms of being retweeted?"
- "Based on the titles, what are the main research topics of COVID-19-related publications? Which research topics attract higher levels of Twitter attention?"
- "On the basis of the full tweet texts and hashtags, what are the topics raised by the Twitter users when discussing the COVID-19-related publications? How does the focus of interest of their Twitter discussions change over time?"


Tracking the Twitter attention around the research efforts on the COVID-19 pandemic

For data about research on COVI-19, data was taken from two different sets: a database of COVID-19 related publications established by the WHO and a file containing COVID-10 related articles, preprints, datasets, and clinical trials from Dimensions.ai. Looking at publications between January 1 ,2020 and March 31, 2020, they extracted 6162 publications for their dataset. 

Altmetric.com was used to extract Twitter data. They looked at Tweets that mentioned publications from their dataset, and found 4195 of their 6162 publications to have Twitter mention data. They extracted 1,374,231 distinct tweets from 655,494 unique Twitter users. Only English tweets were extracted. Information abut the full tweet text, the date posted, the hashtags used, the number of retweets received, the language of the tweet, and profiles of the Twitter users was also extracted. 

Tracking the Twitter attention around the research efforts on the COVID-19 pandemic data collection

Tweets were divided into original vs retweet. Original tweets included original tweets and replies, and retweets included simple retweets and quote tweets. 

VOSviewer was used for text mining and creating co-occurrence networks for terms, descriptions of Twitter users, tweet texts, and hashtags. It was also used to look at the average number of Twitter mentions by term, average number of retweets received y users, and the dates of Tweets posted.

To compare tweets to names of publications, they removed URLs, handle names, the # symbol, and the @ symbol from the tweets, then measured the similarity between these cleaned up tweets and publication names using cosine similarity scores calculated through Python.

Tracking the Twitter attention around the research efforts on the COVID-19 pandemic methods

- Number of research publications and twitter mentions rose from January to March, peaking when Wuhan went into lockdown and when COVID-19 was declared a global outbreak
- a publication abut how COVID-19 is not a lab-manipulated virus got the most Twitter mentions, followed by articles about clinical treatments, public health countermeasures, and transmissibility of the virus
- mentions about articles concentrate a few days after publication 
- Twitter users with an academic or scholarly role are more likely to be retweeted
- Twitter users often tweet their own thoughts and ideas along with information about research when mentioning a publication 
- the focus of discussions has transferred from virological and clinical topics to prevention and treatment, government action, and how the pandemic is affecting society

Learn Before

Related