This paper proposes a new measure of monetary policy surprises based on changes in news media sentiment surrounding monetary policy announcements. Using natural language processing techniques, we analyze news articles covering 194 Monetary Policy Board meetings of the Bank of Korea from 2005 to 2022. Our measure captures policy surprises through news media—the primary information source for households and firms—while allowing economic agents time to process policy information, unlike traditional high-frequency (HF) identification approaches. While HF measures better explain short-term rate movements, our measure contains information across all maturities and generates larger, more persistent impulse responses in macroeconomic variables. More importantly, what differentiates our media-based measure from traditional HF- and VAR-identified shocks is its stronger impact on consumer sentiment and credit spreads. These findings suggest that our media-based approach not only provides a valuable complement to understanding monetary policy transmission channels but also emphasizes the importance of central bank communication.