Flexible, Robust and Dynamic Dialogue Modeling with a Speech Dialogue Interface for Controlling a Hi-Fi Audio System