摘要 |
<p><P>PROBLEM TO BE SOLVED: To objectively and quantitatively evaluate the likelihood of a listener in an interaction system. <P>SOLUTION: In the listener interaction evaluation device 1 for evaluating a listener interaction in the interaction system, an HMM (Hidden Markov Model) learning part 30 creates a listener interaction HMM 40 for modelling the transition state of listener interaction data 10 pre-recorded in the interaction system and creates a chat HMM 50 which models the transition state of chat data 20 pre-recorded in the interaction system. A listener interaction evaluation part 60 applies interaction log data 70 to be evaluated to the listener interaction HMM 40 and the chat HMM 50, calculates output probability of a data sequence of interaction data 70 from the HMM 40, 50, compares output probabilities of both, and evaluates the likelihood of the listener interaction of an interaction 90 to be evaluated. The listener interaction evaluation part 60 evaluates a score based on the logarithm difference of the output probabilities of both and outputs it as numerical evaluation. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |