Evaluating SpeakyAcutattile: A System Based on Spoken Language for Ambient Assisted Living

The current work presents the SPEAKYACUTATTILE system, a computer based platform designed to enable its users to access a number of ICT services such as house management, health status monitoring or recreational services. It is deeply grounded on voice interaction and represents a Spoken Language System (SLS) meant to be a sort of intelligent assistant with which the users can vocally interact. The paper dwells on an experimental assessment of the system. Linear regression analysis have been carried out in order to assess the influence of system performance-in terms of dialogue efficiency metrics and dialogue quality metrics-on the perceived satisfaction of users. Results show that longer dialogues are preferred in order to positively influence the users satisfaction, suggesting the importance for people to have time for getting confident with new ICT solutions. Moreover, from the experimental analysis, it clearly emerges how system intervention should be reduced, suggesting how a good balance between user and system turns represents an important aspect for a fluid interaction. The same can be speculated for dialogues with no met requests that also tended to be associated with lower satisfaction scores. These last results support the importance for a vocal assistant to be effective in its purpose in order to be perceived as useful and satisfying.