BACKGROUND
The novel coronavirus disease (hereafter COVID-19) caused by severe acute respiratory coronavirus 2 (SARS-CoV-2) has caused a global pandemic. During this time, a plethora of information regarding COVID-19 containing both false information (misinformation) and accurate information circulated on social media. The World Health Organization has declared a need to fight not only the pandemic but also the infodemic (a portmanteau of information and pandemic). In this context, it is critical to analyze the quality and veracity of information shared on social media and the evolution of discussions on major topics regarding COVID-19.
OBJECTIVE
This research characterizes risk communication patterns by analyzing public discourse on the novel coronavirus in four Asian countries that suffered outbreaks of varying degrees of severity: South Korea, Iran, Vietnam, and India.
METHODS
We collect tweets on COVID-19 posted from the four Asian countries from the start of their respective COVID-19 outbreaks in January until March 2020. We consult with locals and utilize relevant keywords from the local languages, following each country's tweet conventions. We then utilize a natural language processing (NLP) method to learn topics in an unsupervised fashion automatically. Finally, we qualitatively label the extracted topics to comprehend their semantic meanings.
RESULTS
We find that the official phases of the epidemic, as announced by the governments of the studied countries, do not align well with the online attention paid to COVID-19. Motivated by this misalignment, we develop a new natural language processing method to identify the transitions in topic phases and compare the identified topics across the four Asian countries. We examine the time lag between social media attention and confirmed patient counts. We confirm an inverse relationship between the tweet count and topic diversity.
CONCLUSIONS
Through the current research, we observe similarities and differences in the social media discourse on the pandemic in different Asian countries. We observe that once the daily tweet count hits its peak, the successive tweet count trend tends to decrease for all countries. This phenomenon aligns with the dynamics of the issue-attention cycle, an existing construct from communication theory conceptualizing how an issue rises and falls from public attention. Little work has been performed to identify topics in online risk communication by collectively considering temporal tweet trends in different countries. In this regard, if a critical piece of misinformation can be detected at an early stage in one country, it can be reported to prevent the spread of misinformation in other countries. Therefore, this work can help social media services, social media communicators, journalists, policymakers, and medical professionals fight the infodemic on a global scale.
CLINICALTRIAL
N/A