DeepListener: Harnessing Expected Utility to Guide Clarification Dialog in Spoken Language Systems


Eric Horvitz
Microsoft Research
Redmond, Washington 98052

Tim Paek
Microsoft Research
Redmond, WA 98052


Author email: horvitz@microsoft.com, timpaek@microsoft.com

Abstract:

We describe research on endowing spoken language systems with the ability to consider the cost of misrecognition, and using that knowledge to guide clarification dialog about a user’s intentions. Our approach relies on coupling utility-directed policies for dialog with the ongoing Bayesian fusion of evidence obtained from multiple utterances recognized during an interaction. After describing the methodology, we review the operation of a prototype system called DeepListener. DeepListener considers evidence gathered about utterances over time to make decisions about the optimal dialog strategy or real-world action to take given uncertainties about a user’s intentions and the costs and benefits of different outcomes.

Click here to access postscript or Click here to access pdf format.

Keywords: Bayesian user models, Hidden Markov Model, clarification dialog, joint activity, conversational systems, utility-directed conversation, dialog systems.

In: Proceedings of the 6th International Conference on Spoken Language Processing (ICSLP 2000), Beijing, November 2000.

Related Papers

E. Horvitz and T. Paek, A Computational Architecture for Conversation, Proceedings of the Seventh International Conference on User Modeling, Banff, Canada, June 1999. New York: Springer Wien, pp. 201-210.

E. Horvitz. Uncertainty, Action, and Interaction: In Pursuit of Mixed-Initiative Computing, Intelligent Systems, Sept./ October Issue, IEEE Computer Society.

T. Paek and E. Horvitz, Conversation as Action Under Uncertainty, Proceedings of the 16th Conference on Uncertainty in Artificial Intelligence (UAI-2000), Stanford, CA, June. 2000

T. Paek and E. Horvitz, Uncertainty, Utility, and Misunderstanding: A Decision-Theoretic Perspective on Grounding in Conversational Systems, AAAI Fall Symposium on Psychological Models of Communication in Collaborative Systems, Cape Cod, MA. November 5-7, 1999.

  • T. Paek, E. Horvitz, E. Ringger, Continuous Listening for Unconstrained Spoken Dialog, 6th International Conference on Spoken Language Processing (ICSLP 2000), Beijing, November 2000.