Shaping the Narrative Arc: An Information-Theoretic Approach to Collaborative Dialogue

  • 2019-01-31 18:48:19
  • Kory W. Mathewson, Pablo Samuel Castro, Colin Cherry, George Foster, Marc G. Bellemare
  • 51


We consider the problem of designing an artificial agent capable ofinteracting with humans in collaborative dialogue to produce creative, engagingnarratives. In this task, the goal is to establish universe details, and tocollaborate on an interesting story in that universe, through a series ofnatural dialogue exchanges. Our model can augment any probabilisticconversational agent by allowing it to reason about universe informationestablished and what potential next utterances might reveal. Ideally, with eachutterance, agents would reveal just enough information to add specificity andreduce ambiguity without limiting the conversation. We empirically show thatour model allows control over the rate at which the agent reveals informationand that doing so significantly improves accuracy in predicting the next lineof dialogues from movies. We close with a case-study with four professionaltheatre performers, who preferred interactions with our model-augmented agentover an unaugmented agent.


Introduction (beta)



Conclusion (beta)