We develop a deep learning model to detect emotions embedded in press conferences after the meetings of the Federal Open Market Committee and examine the influence of the detected emotions on financial markets. We find that, after controlling for the Fed's actions and the sentiment in policy...