Development of trainable policies for spoken dialogue systems

• Thesis title in Czech: Vývoj trénovatelných strategií řízení pro dialogové systémy
• Thesis title in English: Development of trainable policies for spoken dialogue systems
• Key words: POMDP, Bayesovské metody, HMM, dialogové systémy, NLP
• English key words: POMDP, Bayesian methods, HMM, dialogue systems, NLP
• Academic year of topic announcement: 2015/2016
• Thesis type: diploma thesis
• Thesis language: angličtina
• Department: Institute of Formal and Applied Linguistics (32-UFAL)
• Supervisor: Mgr. Ing. Filip Jurčíček, Ph.D.
• Author: hidden - assigned and confirmed by the Study Dept.
• Date of registration: 07.12.2015
• Date of assignment: 14.12.2015
• Confirmed by Study dept. on: 17.05.2016
• Date and time of defence: 08.06.2016 09:00
• Date of electronic submission: 16.05.2016
• Date of submission of printed version: 13.05.2016
• Date of proceeded defence: 08.06.2016
• Opponents: Mgr. Nino Peterek, Ph.D.

Guidelines

Recently, a very efficient method for training dialogue policies based on Gaussian processes (GP-SARSA) for spoken dialogue systems was developed. It was shown, that this policy can be trained directly in interaction with real users. The goal of this thesis is to reimplement GP-SARSA and integrate it into the ALEX dialogue systems framework developed at UFAL. The implementation will be tested with simulated and (or) real users. The thesis will asses viability of GP-SARSA for real practical deployments of statistical dialogue systems.

References

C. M. Bishop, Pattern Recognition and Machine Learning, vol. 4, no. 4. Springer, 2006, p. 738.

F. Jurcicek, B. Thomson, S. Young (2011) Reinforcement learning for parameter estimation in statistical spoken dialogue systems. Computer Speech and Language

B. Thomson and S.Young (2010). Bayesian update of dialogue state: A POMDP framework for spoken dialogue systems. Computer Speech and Language.

S. Young, M. Gasic, S. Keizer, F. Mairesse, J. Schatzmann, B. Thomson and K. Yu The Hidden Information State Model: a practical framework for POMDP-based spoken dialogue management. Computer Speech and Language.