Mental State Transition Network (MSTN) isa basic concept of approximating to human psychological and mental responses. It can represent transition from an emotional state to others by a stimulus which Emotion Generating Calculations (EGC) method calculates. In this paper, the agent using Mental State Transition Network can interact with human to realize smooth communicationby an adaptive learning method of the user’s personality trait based mood. The learning method consists of the profit sharing (PS) method and the recurrent neural network (RNN). A sequence of sensor input to MSTN istranslated to an episode which consists of mental state and action. In order to learn the tendency effectively, ineffective rules should be removed from the episode. PSmethod finds out a detour in episode and should be deleted. Furthermore, RNN works to realize the mood according to user’s personality trait. Some experimental results show the variance of human’s delicate emotion.
展开▼