[3S4-IS-4] Robots and real worlds

[3S4-IS-4-03] Generating More Intelligent Responses and Explanations with Conversational AI and Knowledge Graphs

〇Graham Wilcock¹ (1. CDM Interact, Helsinki, Finland / University of Helsinki, Finland)

Regular

Keywords:Knowledge graph, Social robot, Conversational AI, Explainable AI

We have connected social robots to conversational AI systems that search for information in knowledge graphs stored in databases. We use Virtual Furhat robots with Rasa conversational AI. The knowledge graphs are stored in Neo4j graph databases. We have added semantic metadata to the knowledge graphs including taxonomies and other semantic hierarchies extracted from WikiData. Our primary goal is to develop methods to generate more intelligent dialogue responses that leverage the semantic metadata. A further aim is to use the metadata to generate simple explanations if the user asks the robot why it gave a certain response.