Multiparty Verbal Interaction between Humans and Artificial Agents

The study of verbal interaction between multiple humans and robots is an almost unexplored research field. This kind of interaction has been primarily analyzed in the literature focusing on cooperation to achieve a common task or on more technical aspects such as active speaker recognition. The presented work proposes a holistic approach to solve the problem: a cloud architecture that allows social robots and artificial agents to interact verbally with a group of people. The system can recognize the active speaker and decide who to address based on the developed policies while also correctly keeping track of the conversation state.