Weak Signals — Early Warning in Hedged Language

Early warningGlobal · cross-domainCoring: Claim-centricNear-real-time · 4–6× dailyComing soon · vote to prioritise

Hedged language that precedes major events. "Concerns raised by experts," "reports of unusual activity," "preliminary findings suggest" — surfaced as structured claims at Silver tier.

What this dataset is

Major events rarely arrive without hedged language preceding them — expert concern, unusual activity, preliminary findings, strategic reviews. Each weak signal is surfaced as a structured claim at Silver tier, with a precedence score and a link to whatever event eventually materialises. Initial focus: environment, security, finance, health.

Example: a claim, structured

“Concerns raised by experts about elevated activity near the border.”

Becomes a structured Silver record:

signal_type:
Expert concern
domain:
Security
subject:
Cross-border activity
hedge_strength:
high
precedence_score:
0.78

The full per-record property set unlocks with a subscription. Free-tier callers get the main fields shown above plus the geometry needed for the map and graph views.

Source mix at Silver

The Silver gate requires corroboration across independent source categories — multiple outlets in one category don't alone clear the bar.

  • Expert blogs + commentary — Subject-matter expert posts, briefings, hedged analytical commentary.
  • News + wire (hedged language) — Standard news with hedge-detection on phrases like "concerns raised", "preliminary", "unusual".
  • Scientific preprints + working papers — Pre-publication scientific signals across environment, health, security.
  • Regulatory filings with hedged language — Risk-disclosure passages flagging emerging issues in corporate + agency filings.
  • Leaked memos + investigative material — Verified internal documents indicating unannounced concerns.

Per-record fields at Silver

  • · Capture date + date precision
  • · Signal classification (warning / preliminary / hedge / unusual-activity)
  • · Domain (environment / security / finance / health / other)
  • · Originating actor (named expert / agency / outlet)
  • · Subject entity or event the signal points to
  • · Hedge strength (low / medium / high)
  • · Precedence score (model estimate of likelihood the signal precedes a major event)
  • · Link to subsequent confirmed event (if and when materialised)
  • · Source URLs + capture timestamps + content hashes

Free-tier sees a curated subset of these fields in the inspector panel. The full schema (every field above, plus the raw extraction payload) is subscription-only.