摘要 |
<p>A portion of the possibly erroneously recognized words, is marked as current indicated/marked errors using a current measure of confidence. The selection/correction data for selection/correction with respect to the current indicated/marked errors, are received. An improved measure of confidence is determined for remaining erroneously recognized words, for output when a stopping condition is met. Independent claims are also included for the following: (1) speech processing system; (2) computer program product for processing speech; and (3) computer readable storage medium storing speech processing program.</p> |