POMDP (Partially Observable Markov Decision Process) & Dialog Systems


POMDP (Partially Observable Markov Decision Process) & Dialog Systems
Partially Observable Markov Decision Process (POMDP)


Agenda-based user simulation for bootstrapping a POMDP dialogue system [PDF] from upenn.eduJ Schatzmann, B Thomson, K Weilhammer… – … 2007: The Conference …, 2007 – dl.acm.org Abstract This paper investigates the problem of bootstrapping a statistical dialogue manager  without access to training data and proposes a new probabilistic agenda-based method for  simulating user behaviour. In experiments with a statistical POMDP dialogue system, the … Cited by 64 – Related articles – All 19 versions

[PDF] Training and evaluation of the HIS-POMDP dialogue system in noise [PDF] from psu.eduM Gašic, S Keizer, F Mairesse, J Schatzmann… – Proc. 9th SIGdial, 2008 – Citeseer Abstract This paper investigates the claim that a dialogue manager modelled as a Partially  Observable Markov Decision Process (POMDP) can achieve improved robustness to noise  compared to conventional state-based dialogue managers. Using the Hidden Information … Cited by 16 – Related articles – View as HTML – All 18 versions

Bayesian update of dialogue state: A POMDP framework for spoken dialogue systems [PDF] from 134.76.163.171B Thomson… – Computer Speech & Language, 2010 – Elsevier This paper describes a statistically motivated framework for performing real-time dialogue  state updates and policy learning in a spoken dialogue system. The framework is based on  the partially observable Markov decision process (POMDP), which provides a well- … Cited by 35 – Related articles – All 19 versions

Training a real-world POMDP-based Dialogue System [PDF] from upenn.eduB Thomson, J Schatzmann, K Weilhammer… – Proceedings of the …, 2007 – dl.acm.org Abstract Partially Observable Markov Decision Processes provide a principled way to model  uncertainty in dialogues. However, traditional algorithms for optimising policies are  intractable except for cases with very few states. This paper discusses a new approach to … Cited by 19 – Related articles – All 18 versions

Back-off action selection in summary space-based POMDP dialogue systems [PDF] from cam.ac.ukM Gasic, F Lefevre, F Jurcicek… – … , 2009. ASRU 2009. …, 2009 – ieeexplore.ieee.org Abstract This paper deals with the issue of invalid state-action pairs in the Partially  Observable Markov Decision Process (POMDP) framework, with a focus on real-world tasks  where the need for approximate solutions exacerbates this problem. In particular, when … Cited by 7 – Related articles – All 9 versions

A tractable DDN-POMDP approach to affective dialogue modeling for general probabilistic frame-based dialogue systems [PDF] from utwente.nlTH Bui, M Poel, A Nijholt… – 2006 – eprints.eemcs.utwente.nl Abstract We propose a new approach to developing a tractable affective dialogue model for  general probabilistic frame-based dialogue systems. The dialogue model, based on the  Partially Observable Markov Decision Process (POMDP) and the Dynamic Decision … Cited by 11 – Related articles – Library Search – All 18 versions

Effects of user modeling on POMDP-based dialogue systems [PDF] from kaist.ac.krD Kim, HS Sim, KE Kim, JH Kim… – … Annual Conference of …, 2008 – isca-speech.org Partially observable Markov decision processes (POMDPs) have gained significant interest  in research on spoken dialogue systems, due to among many benefits its ability to naturally  model the dialogue strategy selection problem under the unreliability in automated … Cited by 10 – Related articles – All 4 versions

A tractable hybrid DDN-POMDP approach to affective dialogue modeling for probabilistic frame-based dialogue systems [PDF] from utwente.nlTH Bui, M Poel, A Nijholt… – Natural Language …, 2009 – Cambridge Univ Press Abstract We propose a novel approach to developing a tractable affective dialogue model  for probabilistic frame-based dialogue systems. The affective dialogue model, based on  Partially Observable Markov Decision Process (POMDP) and Dynamic Decision Network ( … Cited by 10 – Related articles – All 36 versions

k-nearest neighbor Monte-Carlo control algorithm for POMDP-based dialogue systems [PDF] from rug.nlF Lefevre, M Gašic, F Jurcícek, S Keizer… – Proceedings of the …, 2009 – dl.acm.org Abstract In real-world applications, modelling dialogue as a POMDP requires the use of a  summary space for the dialogue state representation to ensure tractability. Sub-optimal  estimation of the value function governing the selection of system responses can then be … Cited by 6 – Related articles – All 18 versions

Integrating expert knowledge into POMDP optimization for spoken dialog systems [PDF] from att.comJD Williams – AAAI-08 Workshop on Advancements in POMDP …, 2008 – aaai.org Abstract A common problem for real-world POMDP applications is how to incorporate expert  knowledge and constraints such as business rules into the optimization process. This paper  describes a simple approach created in the course of developing a spoken dialog system. … Cited by 6 – Related articles – All 9 versions

[CITATION] Young. S. 2007. Agenda-based user simulation for bootstrapping a POMDP dialog system J Schatzmann, B Thomson, K Weilhammer… – Proceedings of the Annual Conference of … Cited by 3 – Related articles

The hidden information state dialogue manager: A real-world pomdp-based system [PDF] from upenn.eduS Young, J Schatzmann, B Thomson… – Proceedings of Human …, 2007 – dl.acm.org … The Hidden Information State (HIS) dialogue man- ager presented in this demonstration is the first train- able and scalable dialogue system based on the POMDP model. … 2007. Agenda-Based User Simulation for Boot- strapping a POMDP Dialogue System. …  Cited by 2 – Related articles – All 14 versions

[CITATION] POMDP model and its solution for spoken dialogue system B Zhang, QS Cai… – Journal of Computer Research and Development, 2002 Cited by 2 – Related articles

[PDF] D1. 4.1: Improved POMDP-based dialogue system [PDF] from ed.ac.ukS Keizer… – 2009 – wcms.inf.ed.ac.uk Executive summary This document is a short report to accompany the Prototype deliverable  1.4. 1, due at month 12 of the CLASSiC project. It describes the improved version of the  Hidden Information State (HIS) POMDP based spoken dialogue system. A brief overview … Related articles – View as HTML – All 2 versions

[PDF] Uncertainty management for on-line optimisation of a POMDP-based large-scale spoken dialogue system [PDF] from supelec.frL Daubigney, M Gašic, S Chandramohan, M Geist… – 2011 – metz.supelec.fr Abstract The optimization of dialogue policies using reinforcement learning (RL) is now an  accepted part of the state of the art in spoken dialogue systems (SDS). Yet, it is still the case  that the commonly used training algorithms for SDS require a large number of dialogues … Related articles – View as HTML – All 2 versions

Robust Performance Evaluation of POMDP-Based Dialogue Systems D Kim, J Kim… – Audio, Speech, and Language …, 2011 – ieeexplore.ieee.org Abstract Partially observable Markov decision processes (POMDPs) have received  significant interest in research on spoken dialogue systems, due to among many benefits its  ability to naturally model the dialogue strategy selection problem under unreliable … Related articles – All 3 versions

Bayesian reinforcement learning for POMDP-based dialogue systems [PDF] from 140.124.72.88SW Png… – Acoustics, Speech and Signal Processing …, 2011 – ieeexplore.ieee.org Abstract Spoken dialogue systems are gaining popularity with improvements in speech  recognition technologies. Dialogue systems can be modeled effectively using POMDPs,  achieving improvements in robustness. However, past research on POMDPs-based … Related articles – All 6 versions

Divided POMDP method for complex menu problems in spoken dialogue systems M Habibi, S Rahbar… – … Technology Workshop (SLT), …, 2010 – ieeexplore.ieee.org Abstract In this paper, a problem in spoken dialogue systems namely the menu problem, is  introduced and solved by a POMDP model. To overcome the large size of the menu  problem, a new method for achieving an optimal policy called divided POMDP method is … Related articles

[PDF] Optimal dialog in consumer-rating systems using a POMDP framework [PDF] from aclweb.orgZ Li, P Nguyen… – Proceedings of the 9th SIGdial Workshop on …, 2008 – aclweb.org … higher priority. 3 Overview of POMDP 3.1 Basic Definitions A Partially Observable Markov Decision Process (POMDP) is a general framework to handle uncer- tainty in a spoken dialog system. Following nota- 105 Page 3. tions …  Related articles – View as HTML – All 17 versions

[PDF] Semantic graph clustering for POMDP-based spoken dialog systems [PDF] from univ-avignon.frF Pinault… – 2011 – lia.univ-avignon.fr Abstract Dialog managers (DM) in spoken dialogue systems make decisions in highly  uncertain conditions, due to errors from the speech recognition and spoken language  understanding (SLU) modules. In this work a framework to interface efficient probabilistic … Related articles – View as HTML – All 2 versions

[PDF] Unsupervised clustering of probability distributions of semantic frame graphs for POMDP-based spoken dialogue systems with summary space [PDF] from liu.seF Pinault… – … and Reasoning in Practical Dialogue Systems, 2011 – ida.liu.se Abstract Due to errors from the speech recognition and spoken language understanding  modules, dialogue managers in spoken dialogue systems have to make decisions in  uncertain conditions. In this paper a framework to interface efficient probabilistic modeling … Related articles – View as HTML