METHOD AND DEVICE FOR PREDICTING WORD ERROR RATE FROM TEXT

摘要

PROBLEM TO BE SOLVED: To provide a method and a device for predicting a word error rate from a text. SOLUTION: A method of modeling a speech recognition system includes generation of a predicted speech unit sequence by decoding a speech signal generated from a learning text. The learning text includes an actual speech unit sequence used together with the predicted speech unit sequence to form a confusion model. In another embodiment, a text is decoded by using a confusion model and a speech recognition system discriminates the error rate predicted when a speech is decoded on the basis of the text. COPYRIGHT: (C)2004,JPO&NCIPI