This work shows how a spoken dialogue system can be represented as a Partially Observable Markov Decision Process (POMDP) with composite observations consisting of discrete elements representing dialogue acts and continuous components representing confidence scores. Using a testbed simulated dialogue management problem and recently developed optimisation techniques, we demonstrate that this continuous POMDP can outperform traditional approaches in which confidence score is tracked discretely. Further, we present a method for automatically improving handcrafted dialogue managers by incorporating POMDP belief state monitoring, including confidence score information. Experiments on the test-bed system show significant improvements for several example handcrafted dialogue managers across a range of operating conditions.
Description
Partially Observable Markov Decision Processes with Continuous Observations for Dialogue Management | SpringerLink
%0 Book Section
%1 williams2008pomdp
%A Williams, Jason D.
%A Poupart, Pascal
%A Young, Steve
%B Recent Trends in Discourse and Dialogue
%C Dordrecht
%D 2008
%E Dybkjær, Laila
%E Minker, Wolfgang
%I Springer Netherlands
%K decision dialogue management markov observable partially pomdp processes
%P 191--217
%R 10.1007/978-1-4020-6821-8_8
%T Partially Observable Markov Decision Processes with Continuous Observations for Dialogue Management
%U https://doi.org/10.1007/978-1-4020-6821-8_8
%X This work shows how a spoken dialogue system can be represented as a Partially Observable Markov Decision Process (POMDP) with composite observations consisting of discrete elements representing dialogue acts and continuous components representing confidence scores. Using a testbed simulated dialogue management problem and recently developed optimisation techniques, we demonstrate that this continuous POMDP can outperform traditional approaches in which confidence score is tracked discretely. Further, we present a method for automatically improving handcrafted dialogue managers by incorporating POMDP belief state monitoring, including confidence score information. Experiments on the test-bed system show significant improvements for several example handcrafted dialogue managers across a range of operating conditions.
%@ 978-1-4020-6821-8
@inbook{williams2008pomdp,
abstract = {This work shows how a spoken dialogue system can be represented as a Partially Observable Markov Decision Process (POMDP) with composite observations consisting of discrete elements representing dialogue acts and continuous components representing confidence scores. Using a testbed simulated dialogue management problem and recently developed optimisation techniques, we demonstrate that this continuous POMDP can outperform traditional approaches in which confidence score is tracked discretely. Further, we present a method for automatically improving handcrafted dialogue managers by incorporating POMDP belief state monitoring, including confidence score information. Experiments on the test-bed system show significant improvements for several example handcrafted dialogue managers across a range of operating conditions.},
added-at = {2017-09-13T22:49:37.000+0200},
address = {Dordrecht},
author = {Williams, Jason D. and Poupart, Pascal and Young, Steve},
biburl = {https://www.bibsonomy.org/bibtex/228e1a6eba42171eee480755923b5e8c0/porta},
booktitle = {Recent Trends in Discourse and Dialogue},
description = {Partially Observable Markov Decision Processes with Continuous Observations for Dialogue Management | SpringerLink},
doi = {10.1007/978-1-4020-6821-8_8},
editor = {Dybkj{\ae}r, Laila and Minker, Wolfgang},
interhash = {084e4f0d29beb3eefbe851ca74d4122c},
intrahash = {28e1a6eba42171eee480755923b5e8c0},
isbn = {978-1-4020-6821-8},
keywords = {decision dialogue management markov observable partially pomdp processes},
pages = {191--217},
publisher = {Springer Netherlands},
timestamp = {2017-09-13T22:49:37.000+0200},
title = {Partially Observable Markov Decision Processes with Continuous Observations for Dialogue Management},
url = {https://doi.org/10.1007/978-1-4020-6821-8_8},
year = 2008
}