Conversation system for natural communication with robot partner

This paper proposes a conversation system for verbal communication between a human and a robot partner. The system has three main components. In the first one, the conversation system uses the relationship between utterance sentences and perceptual information. Additionally, human utterance patterns are also used to realize a scenario-like conversation. Similarity information between verbs and adjectives of different sentences are also utilized. The second component is the sentence building module which is based on predefined set of rules. Therefore, the robot can also respond using sentence creation. In the third module, the robot can select sentence based on time dependent information. We show experimental results of the proposed method using a real robot partner.