| 摘要 |
PROBLEM TO BE SOLVED: To provide a speech transcribing support device capable of lightening the burden on a correcting operator by accurately detecting a recognition error included in a speech recognition result, and correcting the detected part or presenting a correction example to the correcting operator. SOLUTION: The speech transcribing support device is equipped with a correction instance instructing means 104 for inputting a correct character string to correct a recognition error when the correcting operator finds the recognition error in a recognition result of an input voice, an instance pattern generating means 107 of generating an error instance consisting of a pair of a correct character string for a recognition error and a speech signal of a place corresponding to the recognition error, a similar pattern detecting means 106 of detecting a speech section similar to the speech signal included in the error instance from the input speech, and a recognition result correcting means 105 of replacing a recognition result corresponding to the speech section detected by the similar pattern detecting means 106 with the correct character string, and detects a section wherein a past recognition error instance and a speech signal are similar and replaces the recognition result corresponding to the section with the correct character string. COPYRIGHT: (C)2006,JPO&NCIPI
|
