Data card 09
When is a term in a burst?A question raised by Kleinberg in Bursty and hierarchical structure in streams (2002)
Week to 12 October: “Israel” at four times its expected level; “bullying” and “fentanyl” in no article collected
Articles mentioning each term in the 7 days to 12 October 2026, and the number expected from the 28 days before
“Israel” (以色列) was in 45 collected articles in seven days; at its share of the four weeks before, about 11 were expected. “Bullying” (霸凌) and “fentanyl” (芬太尼), expected in about 11 and 7, were in none.
Source: PropagandaScope, data to 12 October 2026. 25 papers. Expected: each paper’s share of articles mentioning the term in the 28 days to 5 October, applied to that paper’s articles this week, added across the 25 papers (5,619 collected articles in all). The six highest and six lowest ratios among curated terms that pass the site’s thresholds; “export controls” (11 against 1 expected) is left out as too low a baseline. Not Kleinberg’s burst model. Articles cut at 5,000 characters may hide matches.
Research sources: Kleinberg (2002), Bursty and Hierarchical Structure in Streams (PDF)
