公开(公告)号：US20170032790A1

公开(公告)日：2017-02-02

申请号：US15237583

申请日：2016-08-15

Applicant: Intellisist, Inc.

Inventor： Mike O. Webb ,  Bruce J. Peterson ,  Janet S. Kaseda

IPC: G10L15/26 ,  H04M3/533 ,  G10L15/04

CPC classification number: G10L15/26 ,  G10L15/04 ,  H04M3/53333 ,  H04M2201/40 ,  H04M2201/60 ,  H04M2203/2016 ,  H04M2203/4536

Abstract: A computer-implemented system and method for efficient voice transcription is provided. A verbal message is processed by splitting the verbal message into segments and generating text for each of the segments via automated speech recognition. A confidence score is assigned to each text segment. The text segments are provided to workbenches, in order, staring with the text segment having a lowest confidence score. For at least one text segment provided to the workbench, one of edits to the text segment and manually transcribed text to replace the text segment are received. A threshold is applied to a time for performing the message processing and upon satisfaction of the threshold, the message processing is terminated. A text message is generated for the verbal message based on one of the generated text segment, manual transcription, or edited text segment for each of the text segments in that verbal message.

Abstract translation: 提供了一种用于高效语音转录的计算机实现的系统和方法。 通过将言语消息分割成段并通过自动语音识别为每个段生成文本来处理口头消息。 对每个文本段分配置信度分数。 将文本段提供给工作台，以便与具有最低置信度得分的文本段一起凝视。 对于提供给工作台的至少一个文本段，接收文本段的编辑和手动转录文本以替换文本段。 阈值被应用于执行消息处理的时间，并且在满足阈值时，终止消息处理。 基于该语言消息中的每个文本段的生成的文本段，手动转录或编辑的文本段之一，为语言消息生成文本消息。