ForesightHub Trigger Phrases — Multilingual Humanitarian Early-Warning Signals

This dataset contains 60 humanitarian early-warning "trigger phrases" — short text signals extracted from news outlets and public Telegram/Reddit channels across seven world regions. Each record captures a headline or post reporting a potential crisis event (armed conflict, disease outbreak, displacement, state repression, or health-system vulnerability), annotated with the relevant UN/humanitarian agency, an analyst-assigned severity score, a signal-type label, affected country or countries, and source language. Both the original-language text and an English translation are provided. Signals are recorded as reported by their sources and are not independently verified.The dataset was compiled as part of the ForesightHub project, which studies how AI-generated crisis signals are surfaced from open multilingual sources and adopted — or ignored — by humanitarian agencies. It supports research on humanitarian early warning, multilingual NLP, conflict and displacement monitoring, and machine translation of low-resource-language crisis text.Coverage: collected 28 May 2026; 7 regions, 6 agencies, 17 country groupings, 7 languages, 16 sources.