| 摘要 |
PROBLEM TO BE SOLVED: To provide a voice recognizing device which detects an unknown word with high accuracy and can realize unknown word processing that does not affect even the recognition of a part except the unknown word. SOLUTION: An unknown word can be detected with high accuracy by providing a recognition dictionary stored with words information, a language model that has learned each word stored in the recognition dictionary, a recognizing part provided with a phoneme chain model storing a phoneme chain rule and for performing an analysis using the language model and the phoneme chain model, a phoneme chain model weight storing part, and a phoneme chain model weight changing part for changing values stored in the phoneme chain model weight storing part. Further, linguistical constraint can be relieved at unknown word detection and a word following the unknown word can be recognized with high accuracy by providing a language model weight storing part for storing the weight of the language model, and a language model weight changing part for changing the weight stored in the language model weight storing part on the basis of an analysis result obtained by using the language model and the phoneme chain model. COPYRIGHT: (C)2003,JPO
|
