Noah Smith University of Washington Computer Science Any human experience, from dining out to running for the U.S. presidency, prompts us to talk about it. In doing so, we cannot help but select some aspects of the experience to emphasize, at the expense of other aspects. This phenomenon is known as "framing." Framing is perhaps most evident in political communication, where stakes are high, different perspectives on reality lead us to talk past each other, and operatives are accused of spin. In this talk, I will discuss our computational linguistic efforts to grapple with framing, providing an illustrative example for understanding how to build text classifiers.