In plain words
During the COVID-19 outbreak, everyone was worried — but about what, exactly, and in what proportion? Health? Jobs? School closures? This study let the public answer in their own words, at scale.
Pandemic-related posts from Reddit were analysed with LDA (Latent Dirichlet Allocation) topic modelling — an unsupervised technique that groups documents by the themes they share, without anyone pre-deciding what the themes should be. The extracted topics form a map of public concern during the outbreak.
How it works
- Collect the corpusPandemic-related Reddit posts are gathered and cleaned.
- Model the topicsLDA discovers clusters of words that travel together — each cluster a theme of concern.
- Name the themesThe clusters are interpreted into human-readable concerns: health, livelihood, daily-life disruption and more.
- Read the mapThe relative weight of each theme shows what the public actually worried about, over time.
Why it matters
Governments and health agencies communicated constantly during the pandemic, but rarely knew whether they were answering the questions people were actually asking. Topic-modelled social media is a cheap, honest, continuously-updating survey of public concern — useful in any crisis, not just this one.