Handling Conversation Interruption in Many-to-Many HR Interaction Considering Emotional Behaviors and Human Relationships

Abstract

In the future, communication robots are expected to join many-to-many human-robot interactions. Thus, robots must handle interruptions requesting a new task outside of the current conversation. In this paper, we propose a novel scheduler which decides switch timing of conversational tasks when a robot is interrupted. The model grasps the structure of the conversation and finds its breakpoints based on adjacency pairs. In order to decide whether to switch conversational tasks on each breakpoint or not, the model prioritizes conversational tasks considering an importance of its topic and a length as contexts of each conversational task. The model also uses human relationships and emotional behaviors to decide priority of conversational tasks. The result of an evaluation experiment shows that our proposed scheduler could impress subjects more favorably than that which always prioritizes an interrupter.