Date of Award

2010

Publication Type

Master Thesis

Degree Name

M.Sc.

Department

Computer Science

Keywords

Applied sciences

Supervisor

Xiaobu Yuan

Rights

info:eu-repo/semantics/openAccess

Abstract

This thesis proposes a method to manage the interaction between the user and the system dynamically, through speech or text input which updates the user goals, select system actions and calculate rewards for each system response at each time-stamp. The main focus is made on the dialog manager, which decides how to continue the dialogue. We have used POMDP technique, as it maintains a belief distribution on the dialogue states based on the observations over the dialogue even in a noisy environment. Four contextual control modes are introduced in dialogue management for decision-making mechanism, and to keep track of machine behaviour for each dialogue state. The result obtained proves that our proposed framework has overcome the limitations of prior POMDP methods, and exactly understands the actual intention of the users within the available time, providing very interactive conversation between the user and the computer.

Share

COinS