Log-linear dialog manager

Hao Tang, Shinji Watanabe, Tim K. Marks, John R. Hershey

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

We design a log-linear probabilistic model for solving the dialog management task. In both planning and learning we optimize the same objective function: the expected reward. Rather than performing full policy optimization, we perform on-line estimation of the optimal action as a belief-propagation inference step. We employ context-free grammars to describe our variable spaces, which enables us to define rich features. To scale our approach to large variable spaces, we use particle belief propagation. Experiments show that the model is able to choose system actions that yield a high expected reward, outperforming its POMDP-like log-linear counterpart and a hand-crafted rule-based system.

Original languageEnglish
Title of host publication2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages4092-4096
Number of pages5
ISBN (Print)9781479928927
DOIs
Publication statusPublished - 2014
Externally publishedYes
Event2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014 - Florence
Duration: 2014 May 42014 May 9

Other

Other2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014
CityFlorence
Period14/5/414/5/9

Fingerprint

Context free grammars
Knowledge based systems
Managers
Planning
Experiments
Statistical Models

Keywords

  • Dialog Manager
  • Log-linear Model
  • POMDP

ASJC Scopus subject areas

  • Signal Processing
  • Software
  • Electrical and Electronic Engineering

Cite this

Tang, H., Watanabe, S., Marks, T. K., & Hershey, J. R. (2014). Log-linear dialog manager. In 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014 (pp. 4092-4096). [6854371] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP.2014.6854371

Log-linear dialog manager. / Tang, Hao; Watanabe, Shinji; Marks, Tim K.; Hershey, John R.

2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014. Institute of Electrical and Electronics Engineers Inc., 2014. p. 4092-4096 6854371.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Tang, H, Watanabe, S, Marks, TK & Hershey, JR 2014, Log-linear dialog manager. in 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014., 6854371, Institute of Electrical and Electronics Engineers Inc., pp. 4092-4096, 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014, Florence, 14/5/4. https://doi.org/10.1109/ICASSP.2014.6854371
Tang H, Watanabe S, Marks TK, Hershey JR. Log-linear dialog manager. In 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014. Institute of Electrical and Electronics Engineers Inc. 2014. p. 4092-4096. 6854371 https://doi.org/10.1109/ICASSP.2014.6854371
Tang, Hao ; Watanabe, Shinji ; Marks, Tim K. ; Hershey, John R. / Log-linear dialog manager. 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014. Institute of Electrical and Electronics Engineers Inc., 2014. pp. 4092-4096
@inproceedings{713b4d1de7d84e5b87ae07bf90a052e4,
title = "Log-linear dialog manager",
abstract = "We design a log-linear probabilistic model for solving the dialog management task. In both planning and learning we optimize the same objective function: the expected reward. Rather than performing full policy optimization, we perform on-line estimation of the optimal action as a belief-propagation inference step. We employ context-free grammars to describe our variable spaces, which enables us to define rich features. To scale our approach to large variable spaces, we use particle belief propagation. Experiments show that the model is able to choose system actions that yield a high expected reward, outperforming its POMDP-like log-linear counterpart and a hand-crafted rule-based system.",
keywords = "Dialog Manager, Log-linear Model, POMDP",
author = "Hao Tang and Shinji Watanabe and Marks, {Tim K.} and Hershey, {John R.}",
year = "2014",
doi = "10.1109/ICASSP.2014.6854371",
language = "English",
isbn = "9781479928927",
pages = "4092--4096",
booktitle = "2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

TY - GEN

T1 - Log-linear dialog manager

AU - Tang, Hao

AU - Watanabe, Shinji

AU - Marks, Tim K.

AU - Hershey, John R.

PY - 2014

Y1 - 2014

N2 - We design a log-linear probabilistic model for solving the dialog management task. In both planning and learning we optimize the same objective function: the expected reward. Rather than performing full policy optimization, we perform on-line estimation of the optimal action as a belief-propagation inference step. We employ context-free grammars to describe our variable spaces, which enables us to define rich features. To scale our approach to large variable spaces, we use particle belief propagation. Experiments show that the model is able to choose system actions that yield a high expected reward, outperforming its POMDP-like log-linear counterpart and a hand-crafted rule-based system.

AB - We design a log-linear probabilistic model for solving the dialog management task. In both planning and learning we optimize the same objective function: the expected reward. Rather than performing full policy optimization, we perform on-line estimation of the optimal action as a belief-propagation inference step. We employ context-free grammars to describe our variable spaces, which enables us to define rich features. To scale our approach to large variable spaces, we use particle belief propagation. Experiments show that the model is able to choose system actions that yield a high expected reward, outperforming its POMDP-like log-linear counterpart and a hand-crafted rule-based system.

KW - Dialog Manager

KW - Log-linear Model

KW - POMDP

UR - http://www.scopus.com/inward/record.url?scp=84905252896&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84905252896&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2014.6854371

DO - 10.1109/ICASSP.2014.6854371

M3 - Conference contribution

AN - SCOPUS:84905252896

SN - 9781479928927

SP - 4092

EP - 4096

BT - 2014 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2014

PB - Institute of Electrical and Electronics Engineers Inc.

ER -