In plain words
When a crisis hits, does the public mood follow the press — or move on its own? This study put the question to data during COVID-19's first wave, comparing the sentiment of news articles against Twitter from January to July 2020.
To avoid tool bias, three independent sentiment engines were used — TextBlob, VADER and IBM Watson NLU — with cross-tool consistency checks. All three agreed on the pattern: news sentiment hit its most negative in February 2020, while Twitter's negativity peaked in March — correlated, but with the public lagging the press.
How it works
- Build twin corporaPandemic news coverage and pandemic tweets are collected across the same six months.
- Score three waysTextBlob, VADER and IBM Watson each score every text, guarding against any single tool's quirks.
- Align the timelinesMonthly sentiment curves for news and Twitter are laid side by side.
- Read the lagNews negativity peaks in February, the public's in March — evidence that social sentiment follows, not leads.
Why it matters
If public mood trails media tone, then how the press frames a health crisis measurably shapes how frightened — or complacent — the public becomes. The authors argue the same analysis can help readers spot newspaper bias in real time, a media-literacy tool for the next crisis.