Optimizing Dialog Strategies for Conversational Agents Interacting in AmI Environments

In this paper, we describe a conversational agent which provides academic information. The dialog model of this agent has been developed by means of a statistical methodology that automatically explores the dialog space and allows learning new enhanced dialog strategies from a dialog corpus. A dialog simulation technique has been applied to acquire data required to train the dialog model and then explore the new dialog strategies. A set of measures has also been defined to evaluate the dialog strategy. The results of the evaluation show how the dialogmodel deviates from the initially predefined strategy, allowing the conversational agent to tackle new situations and generate new coherent answers for the situations already present in the initial corpus. The proposed technique can be used not only to develop new dialog managers but also to explore new enhanced dialog strategies focused on user adaptation required to interact in AmI environments.