| 摘要 |
Provided is a voice recognition function for integrating in an optimal manner the voice recognition results of a general purpose voice recognition server and the voice recognition results of a dedicated voice recognition server, and which produces few errors in the end. A dedicated voice recognition server (108) is constructed using phrases contained in user dictionary data, and using such data, the functions of a general purpose voice recognition server (106) are evaluated in advance. On the basis of these evaluation results, for recognition results obtained from dedicated and general use voice recognition servers, information related to which results to utilize, and with respect to those results, what weighting to perform thereon to obtain optimal recognition results is preliminarily held as a parameter for result integration (118) in the form of a database. When conducting recognition, optimal recognition results are obtained by comparing recognition results according to dedicated and general purpose voice recognition servers with the parameter for result integration (118). |
