Research Method

We collect the news data from the database The ForSight by Crimson Hexagon. Using search words such as “COVID-19” and “Coronavirus” in different languages, we retrieve all news articles — news titles and lead paragraphs — from major news organizations in different countries and regions.

We use an unsupervised machine learning approach — Latent Dirichlet Allocation topic modeling — to identify 10 main topics from each country’s news coverage by week. Each topic is associated with 20 terms. A group of communication researchers collectively review the terms and decide the labels for each topic. Detailed methods can be found in our previous publication (Guo et al., 2016).