This paper investigates the problem of boot- strapping a statistical dialogue manager with- out access to training data and proposes a new probabilistic agenda-based method for simu- lating user behaviour. In experiments with a statistical POMDP dialogue system, the simu- lator was realistic enough to successfully test the prototype system and train a dialogue pol- icy. An extensive study with human subjects showed that the learned policy was highly com- petitive, with task completion rates above 90%.

Click Here to View the Paper In XML Format