摘要 |
PROBLEM TO BE SOLVED: To carry on a conversation even if a word that a user inputs can not be recognized in a simulated conversation system. SOLUTION: When a voice to a question is inputted, a voice recognition part 210 recognizes the voice (input word) and decides whether there is a registered word to the question. When the input word is already registered, a reply determination part 232 determines a simple reaction word like uh-huh according to the registered word and then determines a next question according to the registered word and current conversation state. If the inputted voice is not recognized or if the voice is recognized, but not the registered word to the question, the replay determination part 232 determines a simple reaction word and a next question according to the conversation state. A voice composition part 220 composes the voices of a simple response word and a next question and outputs the voices through a speaker 30 to carry on the conversation. |