摘要 |
<p>Method for improving the recognition rate of a speech recognition system (4) during a dialogue session with a speaker, comprising the following steps: said speech recognition system (4) verifies at least one predetermined error condition that may cause a deterioration of said recognition rate (208). In a preferred embodiment, the system detects use of unsupported dialects. <??>When one error condition is detected (210), the speech recognition system (4) immediately barges in on the speaker and prompts him to correct the error. <IMAGE></p> |