Abstract
To characterize and compare early coverage of coronavirus disease 2019 (COVID-19) in newspapers, television, and social media, and discuss implications for public health communication strategies that are relevant to an initial pandemic response. Latent Dirichlet allocation (LDA), an unsupervised topic modeling technique, analysis of 3271 newspaper articles, 40 cable news shows transcripts, 96,000 Twitter posts, and 1000 Reddit posts during March 4-12, 2020, a period chronologically early in the timeframe of the COVID-19 pandemic. Coverage of COVID-19 clustered on topics such as epidemic, politics, and the economy, and these varied across media sources. Topics dominating news were not predominantly health-related, suggesting a limited presence of public health in news coverage in traditional and social media. Examples of misinformation were identified, particularly in social media. Public health entities should use communication specialists to create engaging informational content to be shared on social media sites. Public health officials should be attuned to their target audience to anticipate and prevent spread of common myths likely to exist within a population. This may help control misinformation in early stages of pandemics.
Highlights
MethodsTopic Modeling Using Latent Dirichlet allocation (LDA)LDA, an unsupervised machine learning technique,[27] is an exploratory algorithm useful for discovering underlying topics within large bodies of text commonly referred to as a corpus
On December 31, 2019, the World Health Organization (WHO) was alerted to a series of cases of pneumonia of unknown etiology in Wuhan City, China, which were subsequently linked to a seafood and live animal market.[1]
Chinese researchers identified the cause of the disease later named coronavirus disease 2019 (COVID-19) by the WHO1 to be a new type of coronavirus.[2]
Summary
Topic Modeling Using LDALDA, an unsupervised machine learning technique,[27] is an exploratory algorithm useful for discovering underlying topics within large bodies of text commonly referred to as a corpus. LDA has been shown to perform better than other topic modeling techniques in health-related text mining.[29] The goal is to compute the posterior probability given evidence, that is, the conditional distribution of topics, given documents within the corpus.[28] Calculating this requires computation of the joint probability distribution of β, θ, and z across all w (Figure 1) and dividing it by the probability of observing the corpus across all possible topic models. Each word in a document is randomly assigned to a topic, and this process repeats conditioned on the current topic distribution. The algorithm converges when there are no new reassignments, or when the number of iterations is reached, resulting in a per document topic distribution and per topic word distribution for a corpus
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
