Abstract
The previous chapters of this thesis have described statistical methods for policy learning and for handling uncertainty in a spoken dialogue system. Although the parameters are a key component of the statistical models underlying these methods, all experiments thus far have assumed them to be given. ChapterĀ 5 has discussed the optimisation of the policy parameters which describe the policy. This chapter will discuss the optimisation of the parameters of the Bayesian network used for user modeling. These include the probabilities for changes in user goals and the probability of user actions given the goals.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Note that the \(H_{c,u}\) sets are dependendent on the machine act, the concept, and user act.
- 2.
The Cartesian product, \(H_u=\prod\nolimits _c H_{c,u}\) is the set \(\left\{ \mathbf{g }: \forall k, g_k \in H_{u,k} \right\} \).
- 3.
The given equations do not take into account the possibility of division by zero. This is ignored here, but can easily be included without any extra computational burden.
- 4.
Note that the additional 216 dialogues were recorded using a separate Hidden Information State dialogue manager not described here.
- 5.
The separate trial used different subjects and different task specifications.
References
Bernardo J (1976) Algorithm AS 103: Psi (Digamma) function. In: Applied statistics, vol 25 of 3. Blackwell Publishing for the Royal Statistical Society, pp 315ā317
Doshi F, Roy N (2007) Efficient model learning for dialog management. In: Proceedings of the ACM/IEEE international conference on human-robot interaction. ACM, pp 65ā72. ISBN 978-1-59593-617-2
Keizer S, Gasic M, Mairesse F, Thomson B, Yu K, Young S (2008) Modelling user behaviour in the HIS-POMDP dialogue manager. In: Proceedings of SLT, pp 121ā124
Minka T (2001) A family of algorithms for approximate Bayesian inference. Ph.D. thesis, MIT Press
Paquet U (2007) Bayesian inference for latent variable models. Ph.D. thesis, University of Cambridge
Schneider B (1978) Algorithm AS 121: trigamma function. In: Applied Statistics, vol 27 of 1. Blackwell Publishing for the Royal Statistical Society, pp 97ā99
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
Ā© 2013 Springer-Verlag London
About this chapter
Cite this chapter
Thomson, B. (2013). Parameter Learning. In: Statistical Methods for Spoken Dialogue Management. Springer Theses. Springer, London. https://doi.org/10.1007/978-1-4471-4923-1_7
Download citation
DOI: https://doi.org/10.1007/978-1-4471-4923-1_7
Published:
Publisher Name: Springer, London
Print ISBN: 978-1-4471-4922-4
Online ISBN: 978-1-4471-4923-1
eBook Packages: EngineeringEngineering (R0)