NTU-led study shows global sentiment on social media towards Covid-19 has shifted from fear to anger

"Word clouds" generated by a machine-learning algorithm used by the NTU team.
"Word clouds" generated by a machine-learning algorithm used by the NTU team.PHOTO: NANYANG TECHNOLOGICAL UNIVERSITY

SINGAPORE - People's feelings about Covid-19 around the world have shifted from fear to anger and joy as the pandemic developed, according to a study led by Nanyang Technological University (NTU) that analysed 20 million tweets on Twitter from January to April.

The team's findings released on Wednesday (July 15) showed that fear was the dominant emotion from late January to early March, accounting for up to half of tweets analysed daily.

But that began to decline as anger spiked in late February, peaking at 29 per cent of daily tweets on March 12, a day after the World Health Organisation declared the Covid-19 outbreak a pandemic, while joy began to climb slowly from late March to early April.

The study analysed tweets in English that made reference to the coronavirus. A machine-learning algorithm called CrystalFeel developed by the Agency for Science, Technology and Research (A*Star) was used to identify joy, sadness, fear and anger based on commonly occurring words or phrases.

For example, words such as "scared" or "panic" could have appeared alongside Covid-19 related terms such as "confirmed case" or "test shortages". These point to fears over the initial unknowns surrounding the coronavirus and later the shortage of Covid-19 tests, the team said.

On the other hand, xenophobia was a common theme among anger-related tweets at the start of the pandemic when China was the epicentre of the outbreak, with "Chinese people" one of the most common phrases found.

But the anger subsequently evolved in April, the team said, with "stay home" and a variety of swear words becoming more common in tweets, reflecting negative feelings associated with isolation and social seclusion.

The study also found an uptick in sentiments of joy relating to national pride and community spirit as time went on, which researchers said offered a "glimmer of hope".

"Our findings suggest that collective issues driven by emotions, such as shared experiences of distress (during) the Covid-19 pandemic... are developing," said team leader Professor May O. Lwin, who is from NTU's Wee Kim Wee School of Communication and Information.

China's Tianjin University, the University of Lugano in Switzerland, and Australia's University of Melbourne were among the collaborators in this study.

Prof Lwin added that widespread mistrust and a belief in online falsehoods could hinder the ongoing control of the disease, should such emotions not be addressed through clear and decisive communication by governments, citizen groups and social media stakeholders.

A graph showing evolving public emotions over the course of the Covid-19 pandemic from January to June based on tweets analysed by an NTU-led team. PHOTO: NANYANG TECHNOLOGICAL UNIVERSITY

More recent data from May and June continued to reflect an increasing trend of anger in tweets, while fear became less common than both anger and joy.

Sadness has been significantly lower than the other three emotions throughout the year, although the study did not specify why that might have been the case.


Prof Lwin intends to conduct follow-up studies to examine country-specific trends in public emotions.

Preliminary findings indicate that there has been a moderate balance of positive sentiments in Singapore, relating to resilience, civic pride and celebration of acts of kindness

This is in contrast to other countries where strong negative emotions overwhelmingly feature in social media posts, the researchers said.

The study is funded by A*Star and the National Research Foundation under the Covid-19 Research Fund administered by the Ministry of Health.