This thesis is part of a multidisciplinary context on the border between the domains of the ambient intelligence, the human-machine interaction and the conversational agents. We study the problem of contextual proactive assistance in the course of the ambient computing. In this regard, our challenge is to provide to the user who wish have a virtual assistant, a proactive services which can communicate according to the user's context. However proactive ambient intelligence, or contextual proactive assistance, did not follow an important development as reactive ambient intelligence where the user must explicitly express their needs. This is why the objective of this CIFRE thesis is to study and evaluate the concept of contextual proactive assistance through the implementation of an online conversational agent named X-CAMPUS. This acronym stands for eXtensible Proactive Multichannel Conversational Agent for Ubiquitous Services. This system aims to assist the user in their daily tasks thanks to its ability to perceive the state of the environment and interact effectively according to user's needs (weather, TV programs, etc.). We conducted an experimental study to investigate the interest of X-CAMPUS and evaluate the satisfaction that it provides to users. We describe quantitative and qualitative results of this study obtained through a set of scenarios using different parameters of interaction (multimodal, multi-channels, behavior reactive/proactive agent, type of sensors, context mono/multi-user…).
Directeur de Thèse : José ROUILLARD Co-encadrant de Thèse : Jean-Claude TARBY Rapporteurs : Emmanuel DUBOIS, Frédérique LAFOREST Membres : Carole ADAM, Isabelle PECCI, Jean-Christophe ROUTIER, François BRICOUT
Thesis of the team NOCE defended on 19/11/2013