Video Recordings

Ever wondered why people interrupt each other? It is not just because they are impolite. Often this is because hearers understand what speakers want to say before the end of the utterance. That is, the meaning of the utterance becomes actionable before the speaker has finished.  Actionability-based language understanding shifts the overall objective of computational linguistics from deriving complete semantic and discourse/pragmatic interpretations of language inputs to understanding just enough to confidently maintain a conversation, learn from a text, or participate in a team task.  In this talk, we will motivate and illustrate the methodological shift toward actionability by considering the desiderata for developing artificial intelligent agents that are human-like in their abilities to understand, learn, and explain.