The EEE corpus: socio-affective “glue ” cues in elderly-robot interactions in a Smart Home with the EmOz platform

摘要

The aim of this preliminary study of feasibility is to give a glance at interactions in a Smart Home prototype between the elderly and a companion robot that is having some socio-affective language primitives as the only vector of communication. The paper particularly focuses on the methodology and the scenario made to collect a spontaneous corpus of human-robot interactions. Through a Wizard of Oz platform (EmOz), which was specifically developed for this issue, a robot is introduced as an intermediary between the technological environment and some elderly who have to give vocal commands to the robot to control the Smart Home. The robot vocal productions increases progressively by adding prosodic levels: (1) no speech, (2) pure prosodic mouth noises supposed to be the “glue’s ” tools, (3) lexicons with supposed “glue ” prosody and (4) subject’s commands imitations with supposed “glue ” prosody. The elderly subjects ’ speech behaviours confirm the hypothesis that the socio-affective “glue ” effect increase towards the prosodic levels, especially for socio-isolated people. The actual corpus is still on recording process and is motivated to collect data from socio-isolated elderly in real need.