摘要 |
PROBLEM TO BE SOLVED: To provide a voice recognition device that can evaluate a learning result of the speaker to input the voice by observing the responses of the speaker when there occur recognition errors of the inputted voice. SOLUTION: When the voice input is repeated due to recognition errors, the history showing "Another voice input has been made by canceling the first voice input" or " Another voice input has been made by canceling all the voice input" and the time required for another inputting are recorded in the history management section 110 matching with the recognition errors. The content of the above history management section 110 is analyzed by the learning result evaluation section 111 checking how often the voice input has been repeated after going back one voice input, how often the voice input has been repeated after canceling all the previous voice inputs, and the time required for repeating the input. Thus the learning result of the speaker is evaluated. COPYRIGHT: (C)2005,JPO&NCIPI
|