A Hybrid Conversational Agent with Semantic Association of Autobiographic Memories for the Elderly
Journal
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Journal Volume
12193 LNCS
Pages
53-66
Date Issued
2020
Author(s)
Abstract
Socially Assistive Robots are becoming essential in the field of elderly care, as they can support caregivers in their tasks, for instance, by providing senior users with emotional and psychological support through verbal communication. In this paper, we present the results of a project where we developed an interactive dialogue system so that a robot could engage elderly users in conversations about their personal life stories. A task that seems almost mundane for the average person, is in fact extremely challenging for a machine to achieve. Through the development of a comprehensive platform with a variety of modules, the system is able to extract essential keywords from a user utterance and classify them according to sentence context and word meaning. These keywords are then indexed in a user-specific knowledge base, where semantic associations between items are made, relating them, for instance, by time or place. These items are used by the robot to generate responses to the user’s speech by leveraging a hybrid template/data-driven mechanism. As the user interacts with the system, it learns more details which further enrich the generated sentences. The system was evaluated on a human-in-loop experiment, where the results showed the ability of the system to understand human speech, memorize personal information of each user and generate coherent responses in dialogic interactions. These results highlight the potential of a robot not only to provide companionship, but also to build a social relationship with its user.
SDGs
Type
conference paper