The Loupe Digger

Transparency

How these numbers are made

We analyse headlines and ledes from major news outlets. Our key metric is emotional density: the number of emotionally loaded words divided by the total words in each headline and lede, expressed as a percentage. For example, two loaded words in 40 words gives 5%. Digger averages percentages across articles, so each article counts equally, regardless of length. We measure the newsroom’s framing, not the feelings it causes in a reader.

The editor — a professional reporter — reviews each article. AI assists in highlighting emotional words and phrases in the headline and lede in context and assigning emotion labels in line with The Loupe Codebook principles. The editor double checks findings, removes inaccurate matches, adds missed wording and corrects emotion labels against the same codebook.

The Loupe Codebook and counting rules

Seven emotions

  • Anger — outrage, confrontation, blame, aggression (blasted, slammed, furious, retaliation)
  • Fear — threat, danger, vulnerability, looming crisis (crisis, spiral, panic, looms)
  • Sadness — loss, grief, tragedy, devastation (mourns, heartbreak, crushed, tragedy)
  • Disgust — moral aversion, corruption, betrayal (scandal, corrupt, shameful, smear)
  • Surprise — shock, abrupt disruption, the unexpected (bombshell, stunned, abrupt)
  • Joy / Triumph — victory, celebration, dominance (triumph, jubilant, hailed, sweep)
  • Hope / Relief — de-escalation, recovery, optimism, reassurance (ceasefire, breakthrough, eases, lifeline)
  • Intensifier (additional) — amplifies scale, force or certainty without expressing a specific emotion. Counts towards emotional density, but is not an eighth emotion. (massive, unprecedented, sharply)

Editorial principles

  1. Sentence context. The same word can carry different emotions or none. “Talks collapse” is Fear (threat of breakdown); “bridge collapses after storm” is a plain fact and is not counted.
  2. Dominant emotion. Where a phrase touches several emotions (“shameful betrayal”), the editor picks the main tone of the angle.
  3. Article weight. Each article counts equally when calculating emotional density, reflecting how a reader perceives individual stories, not all pooled coverage.
  4. Ranking threshold. Most- and least-emotional rankings include only groups with at least three articles.

Confidence labels

  • High — 8+ articles and 300+ analysed words.
  • Medium — 3+ articles and 100+ analysed words.
  • Low — anything thinner. Treat as indicative only.

Copyright and quotation

Digger analyses only headlines and ledes, quoted for criticism and review under the UK's fair dealing rules, credited to each outlet and linked to the original. Copyright remains with the publishers; we never reproduce full articles or bypass paywalls.

Known limitations
  • We analyse headline + lede, not full articles. A few loaded words can move a single article’s density substantially.
  • English only; the corpus is currently all English-language reporting.
  • We measure emotional framing in the wording, not readers’ feelings, an outlet’s intentions or political bias. A dramatic event legitimately produces dramatic vocabulary.