A computational model for automatic generation of domain-specific dialogues using machine learning

Abstract

Automatic generation of dialogues is a very important component for the human-robot interaction task ; the dialogues generated must guarantee a coherent conversation between human beings and robots. The aim is that the interaction is as natural and effective as possible, considering aspects such as: age, gender, socio-cultural level, socio-economic level, and so on. This research report presents an overview of a doctoral research work that is intended to be executed during the following three years. We motivate the necessity of researching in automatic generation and evaluation of dialogues in the framework of human-robot interaction, presenting the related work reported in literature, the research objectives and the methodology we are interested in develop. In general, we propose to employ machine learning techniques in a restricted domain of knowledge for generating the human-robot dialogues.