On-Line Learning of a Persian Spoken Dialogue System Using Real Training Data

The first spoken dialogue system developed for the Persian language isintroduced. This is a ticket reservation system with Persian ASR and NLU modules.The focus of the paper is on learning the dialogue management module. In thiswork, real on-line training data are used during the learning process. For on-linelearning, the effect of the variations of discount factor (g) on the learning speed isinvestigated as the second contribution of the research. The optimal values for gwere found and the variation pattern of the action-value function (Q) in the learningprocess was obtained. A probabilistic policy for selecting actions is used in thiswork for the first time instead of greedy policies employed in previous works

Keywords:

Learning spoken dialogue , action value function , discount factor