Domain Metric Knowledge Model for Embodied Conversation Agents

Abstract — In this paper, we propose a domain metric knowledge model for Embodied Conversational Agents (ECAs) for human computer interaction. In our research, we have found that by incorporating Open-Domain and Domain-Specific knowledge, a new generation of ECAs will be able to demonstrate intelligence and to increase trust by the users. The proposed system is aimed at advancing the scholarly field of artificial intelligence and to offer assistance to queries on real-world issues. The experimental system we have developed aims to provide answers to questions on pandemic threat. The focus is on Bird Flu as the Domain-Specific knowledge together with the support of multiple open-domain knowledge bases. We also present a scheme for selecting specific websites as the input knowledge bases for our conversational system based on reputability, creditability, reliability and trustworthiness