Cost-level integration of statistical and rule-based dialog managers

Shinji Watanabe, John R. Hershey, Tim K. Marks, Youichi Fujii, Yusuke Koji

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Statistical dialog managers can potentially make more robust decisions than their rule-based counterparts, because they can account for uncertainties due to errors in speech recognition and natural language understanding. In practice, however, statistical dialog managers can be difficult to use, as they may require a large number of parameters to be inferred from limited data. Consequently, hand-crafted rule based systems are still effective for practical use. This paper proposes a method to integrate an existing rule-based dialog manager with a statistical dialog manager based on Bayes decision theory, by incorporating the rule-based dialog manager into the cost function of the statistical dialog manager. The cost function has two parts: An efficiency cost that penalizes inefficient actions, as in conventional statistical dialog approaches, and a regularization cost that slightly penalizes system actions that differ from those that would be chosen by the rule-based system. Our experiments, which use a destination-setting task in an automobile dialog scenario, demonstrate that the integrated system produces system actions that are similar to those of an existing rule-based dialog manager but enable task completion using fewer turns than the rule-based system.

Original languageEnglish
Pages (from-to)323-327
Number of pages5
JournalUnknown Journal
Publication statusPublished - 2014
Externally publishedYes

Fingerprint

Managers
costs
Costs
Knowledge based systems
Rule-based Systems
Cost functions
decision theory
Cost Function
Decision theory
automobiles
speech recognition
Dialogue
Speech recognition
Decision Theory
Automobiles
Automobile
Bayes
Integrated System
Speech Recognition
Natural Language

Keywords

  • Cost-level integration
  • Goal estimation
  • Rule based system
  • Spoken dialog system
  • Statistical dialog manager

ASJC Scopus subject areas

  • Language and Linguistics
  • Human-Computer Interaction
  • Signal Processing
  • Software
  • Modelling and Simulation

Cite this

Watanabe, S., Hershey, J. R., Marks, T. K., Fujii, Y., & Koji, Y. (2014). Cost-level integration of statistical and rule-based dialog managers. Unknown Journal, 323-327.

Cost-level integration of statistical and rule-based dialog managers. / Watanabe, Shinji; Hershey, John R.; Marks, Tim K.; Fujii, Youichi; Koji, Yusuke.

In: Unknown Journal, 2014, p. 323-327.

Research output: Contribution to journalArticle

Watanabe, S, Hershey, JR, Marks, TK, Fujii, Y & Koji, Y 2014, 'Cost-level integration of statistical and rule-based dialog managers', Unknown Journal, pp. 323-327.
Watanabe, Shinji ; Hershey, John R. ; Marks, Tim K. ; Fujii, Youichi ; Koji, Yusuke. / Cost-level integration of statistical and rule-based dialog managers. In: Unknown Journal. 2014 ; pp. 323-327.
@article{51920df9dba54702bbe4c250e4032749,
title = "Cost-level integration of statistical and rule-based dialog managers",
abstract = "Statistical dialog managers can potentially make more robust decisions than their rule-based counterparts, because they can account for uncertainties due to errors in speech recognition and natural language understanding. In practice, however, statistical dialog managers can be difficult to use, as they may require a large number of parameters to be inferred from limited data. Consequently, hand-crafted rule based systems are still effective for practical use. This paper proposes a method to integrate an existing rule-based dialog manager with a statistical dialog manager based on Bayes decision theory, by incorporating the rule-based dialog manager into the cost function of the statistical dialog manager. The cost function has two parts: An efficiency cost that penalizes inefficient actions, as in conventional statistical dialog approaches, and a regularization cost that slightly penalizes system actions that differ from those that would be chosen by the rule-based system. Our experiments, which use a destination-setting task in an automobile dialog scenario, demonstrate that the integrated system produces system actions that are similar to those of an existing rule-based dialog manager but enable task completion using fewer turns than the rule-based system.",
keywords = "Cost-level integration, Goal estimation, Rule based system, Spoken dialog system, Statistical dialog manager",
author = "Shinji Watanabe and Hershey, {John R.} and Marks, {Tim K.} and Youichi Fujii and Yusuke Koji",
year = "2014",
language = "English",
pages = "323--327",
journal = "Nuclear Physics A",
issn = "0375-9474",
publisher = "Elsevier",

}

TY - JOUR

T1 - Cost-level integration of statistical and rule-based dialog managers

AU - Watanabe, Shinji

AU - Hershey, John R.

AU - Marks, Tim K.

AU - Fujii, Youichi

AU - Koji, Yusuke

PY - 2014

Y1 - 2014

N2 - Statistical dialog managers can potentially make more robust decisions than their rule-based counterparts, because they can account for uncertainties due to errors in speech recognition and natural language understanding. In practice, however, statistical dialog managers can be difficult to use, as they may require a large number of parameters to be inferred from limited data. Consequently, hand-crafted rule based systems are still effective for practical use. This paper proposes a method to integrate an existing rule-based dialog manager with a statistical dialog manager based on Bayes decision theory, by incorporating the rule-based dialog manager into the cost function of the statistical dialog manager. The cost function has two parts: An efficiency cost that penalizes inefficient actions, as in conventional statistical dialog approaches, and a regularization cost that slightly penalizes system actions that differ from those that would be chosen by the rule-based system. Our experiments, which use a destination-setting task in an automobile dialog scenario, demonstrate that the integrated system produces system actions that are similar to those of an existing rule-based dialog manager but enable task completion using fewer turns than the rule-based system.

AB - Statistical dialog managers can potentially make more robust decisions than their rule-based counterparts, because they can account for uncertainties due to errors in speech recognition and natural language understanding. In practice, however, statistical dialog managers can be difficult to use, as they may require a large number of parameters to be inferred from limited data. Consequently, hand-crafted rule based systems are still effective for practical use. This paper proposes a method to integrate an existing rule-based dialog manager with a statistical dialog manager based on Bayes decision theory, by incorporating the rule-based dialog manager into the cost function of the statistical dialog manager. The cost function has two parts: An efficiency cost that penalizes inefficient actions, as in conventional statistical dialog approaches, and a regularization cost that slightly penalizes system actions that differ from those that would be chosen by the rule-based system. Our experiments, which use a destination-setting task in an automobile dialog scenario, demonstrate that the integrated system produces system actions that are similar to those of an existing rule-based dialog manager but enable task completion using fewer turns than the rule-based system.

KW - Cost-level integration

KW - Goal estimation

KW - Rule based system

KW - Spoken dialog system

KW - Statistical dialog manager

UR - http://www.scopus.com/inward/record.url?scp=84910048701&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84910048701&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84910048701

SP - 323

EP - 327

JO - Nuclear Physics A

JF - Nuclear Physics A

SN - 0375-9474

ER -