Central bank speak, decoded: 225k sentences stance-labeled across 26 banks
A new CC-BY-4.0 corpus crawls monetary-policy communications from 26 central banks and labels each sentence for policy stance, sentiment, topic, and doc type — 225,101 annotations spanning 1995 to 2026. It covers 21 languages with English translations, 38k documents, and a linked dashboard. Anyone building models to read the language of central banks gets a rare open, ready-to-train dataset.