DISCRIMINATIVE VOICE RECOGNITION ACCURACY ESTIMATION DEVICE, DISCRIMINATIVE VOICE RECOGNITION ACCURACY ESTIMATION METHOD, AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide a discriminative voice recognition accuracy estimation device capable of obtaining an estimation value of recognition accuracy with high accuracy.SOLUTION: A discriminative voice recognition accuracy estimation device includes: a voice recognition unit for voice-recognizing an input voice so as to output a voice recognition result; a word feature amount vector creation unit for creating a word feature amount vector on the basis of information obtained from the voice recognition result; and a recognition result word stochastic classification unit for creating a probability word classification result expressing, for each segment, a probability of a word containing &egr; in a 1-best recognition result word string being a correct answer, a probability of a substitution error, a probability of an insertion error, and a probability of a deletion error by using a word feature amount vector for each word containing &egr; in a word string (1-best recognition word string) created connecting words having a maximum existence probability in a segment, and a recognition result word stochastic classification model previously learned. The voice recognition accuracy is calculated by using the stochastic word classification result.