The fear that people developed at the start of the COVID-19 outbreak has given way to anger over the course of the pandemic, a study of global sentiments led by NTU Singapore with IHPC’s algorithm to analyse underlying emotions of tweets in English has found.

To identify trends in the expression of the four basic emotions – fear, anger, sadness, and joy – and examine the narratives underlying those emotions, the team collected 20,325,929 tweets in English containing the keywords ‘Wuhan’, ‘corona’, ‘nCov’, and ‘covid’.

The tweets, collected from late January to early April by IHPC using Twitter’s standard search application interface programme, came from over 7 million unique users in more than 170 countries.

The underlying emotions of tweets were then analysed using a machine-learning algorithm called CrystalFeel developed by IHPC whose accuracy has been demonstrated in previous studies. Word clouds based on the top single words and two-word phrases were generated for each of the four emotions. Upon analysing the results, the team found that words such as ‘first case’ and ‘outbreak’ were among the most-used words in tweets from late January, indicating fear that was possibly related to the emerging coronavirus and the unknown nature of it, causing uncertainty about containment and spread.

The study was published in the scientific journal JMIR Public Health & Surveillance in May and the global team involved in the study includes IHPC, NTU Singapore, Tianjin University, University of Lugano, and University of Melbourne.

Read about the study featured in the media here.