Using utterance-level confidence estimates

摘要: A computer-based method of routing a message to system includes receiving message, and processing the using large-vocabulary continuous speech recognition generate string text corresponding message. The generating confidence estimate comparing predetermined threshold. If satisfies threshold, is forwarded system. does not satisfy information relating transcriptionist. may include one or more utterances. Each utterance in be separately jointly processed. In this way, generated evaluated for each whole Information based on results generation evaluation.