Global Sentiments Surrounding the COVID-19 Pandemic on Twitter: Analysis of Twitter Trends

Public psychological wellbeing has emerged as a clear area of concern as the negative impact of the COVID-19 pandemic grow with time, and needs to be managed for societies to minimise negative outcomes and function as normally as possible. Aided by the high proportion of social media users and relative ease of obtaining large amounts of social media data, analysis of the emotions expressed on social media and the underlying narratives could provide quick actionable insights for management of public psychological wellbeing.  

This study analysed 20.3 million tweets sent by more than 7 million unique users spanning more than 170 countries in 28 Jan 20 to 9 Apr 20 containing the terms "wuhan", "corona", "nCov" and "covid" – terms widely used at various points in time to refer to COVID-19 – for expressions of four emotions: fear, anger, sadness and joy. The analysis, carried out using the CrystalFeel sentiment analysis platform developed in-house by IHPC, captured the changes in global sentiment with regard to COVID-19.

  • Fear was expressed in about 50% of the tweets sent in end-January when the disease first surfaced, but gradually dropped as the crisis developed to less than 30% of tweets in early-April.
  • In contrast, tweets expressing anger – an emotion in response to external uncertainties – progressively increased from late-January to early-March, peaking at 29% on 12 Mar 20, a day after the pandemic declaration by WHO.
  • Tweets expressing sadness such as that resulting from friends or family members passing on, although proportionally lower than those of the other emotions, doubled since the WHO declaration.
  • Similarly, tweets expressing joy, suggesting a sense of pride, gratitude, hope, and happiness, also increased.

  • Further analyses using word clouds illuminated the evolving narratives underlying these emotions.  For instance, fear was mainly associated with the uncertainty of COVID-19 containment and spread in end-January, but evolved to include concerns regarding shortages of COVID-19 tests and medical supplies; and while anger was initially predominantly localised to China and Asia, anger shifted to include the inadvertent consequences of safe distancing measures, such as isolation fatigue and social seclusion from prolonged stays at home.

    The prevalence of negative emotions and the rapid evolution of sentiments revealed in this study corroborates the need for careful management of the public's mental wellbeing in this unprecedented crisis.  The international collaboration team (NTU, Tianjin University, University of Lugano, University of Melbourne and IHPC) has also investigated differences in sentiment among countries such as Singapore, United States, South Korea and Italy. Since the completion of these two rapid reports, IHPC has worked on preparing the worldwide Twitter data tagged with our in-house algorithms processed topics, sentiments and emotions attributes, and is planning to make it available to the broader research community as a research resource for further investigations. 

    IHPC has also engaged MOH Office of Health Transformation regarding a national ground sensing of public mental health status using Facebook data. Longer-term research questions include: deeper analyses of the cultural accounts of the various emotional responses, analysis of different media's coverage and stances in covering COVID-19 related news, and investigation of how the social media data stream can provide a predictive layer in epidemiological science in the surveillance of potential new infectious disease threats.

    The article titled "Global Sentiments Surrounding the COVID-19 Pandemic on Twitter: Analysis of Twitter Trends" was published in JMIR Public Health and Surveillance (impact factor not available) by researchers from NTU, Tianjin University, University of Lugano, University of Melbourne and IHPC.  The IHPC researchers involved in this work are Dr Raj Kumar Gupta and Dr Yang Yinping.