Prosodic analysis in human-machine interaction

Abstract

In this paper, we are going to present some experiments concerning the analysis of prosodic features in the spoken production of requests by human users in human-machine interactions. The main aim of this analysis is to understand if and how much a speaker adapts the spoken production to his/her virtual interlocutor, and to which extent this could be caused by the type of user and his/her representational preconception towards the specific interaction. The collected results are, therefore, considered as an important means for developing spoken dialogue systems whose speech recognition module skills are better suited to the characteristic of the human interlocutor.