-
公开(公告)号：US20170185286A1
公开(公告)日：2017-06-29
申请号：US14982887
申请日：2015-12-29
Applicant: Google Inc.
Inventor： Francoise Beaufays , Yu Ouyang , David Rybach , Michael D. Riley , Lars Hellsten
IPC: G06F3/0488 , G06F3/01 , G06F3/041
CPC classification number: G06F3/04886 , G06F3/017 , G06F3/0237 , G06F3/0416 , G06F2203/04106
Abstract: Methods, systems, and apparatus for receiving data indicating a location of a particular touchpoint representing a latest received touchpoint in a sequence of received touchpoints; identifying candidate characters associated with the particular touchpoint; generating, for each of the candidate characters, a confidence score; identifying different candidate sequences of characters each including for each received touchpoint, one candidate character associated with a location of the received touchpoint, and one of the candidate characters associated with the particular touchpoint; for each different candidate sequence of characters, determining a language model score and generating a transcription score based at least on the confidence score for one or more of the candidate characters in the candidate sequence of characters and the language model score for the candidate sequence of characters; selecting, and providing for output, a representative sequence of characters from among the candidate sequences of characters based at least on the transcription scores.
-
公开(公告)号：US20150127327A1
公开(公告)日：2015-05-07
申请号：US14282655
申请日：2014-05-20
Applicant: Google Inc.
Inventor： Michiel A.U. Bacchiani , David Rybach
IPC: G10L25/30 , G10L19/038 , G10L15/26
Abstract: The technology described herein can be embodied in a method that includes receiving an audio signal encoding a portion of an utterance, and providing, to a first neural network, data corresponding to the audio signal. The method also includes generating, by a processor, data representing a transcription for the utterance based on an output of the first neural network. The first neural network is trained using features of multiple context-dependent states, the context-dependent states being derived from a plurality of context-independent states provided by a second neural network.
Abstract translation: 本文描述的技术可以包括接收编码话音的一部分的音频信号并向第一神经网络提供对应于音频信号的数据的方法。 该方法还包括基于第一神经网络的输出，由处理器生成表示用于话语的转录的数据。 使用多个上下文相关状态的特征训练第一神经网络，所述上下文相关状态从由第二神经网络提供的多个与上下文无关的状态导出。
-
公开(公告)号：US09620145B2
公开(公告)日：2017-04-11
申请号：US14282655
申请日：2014-05-20
Applicant: Google Inc.
Inventor： Michiel A. U. Bacchiani , David Rybach
Abstract: The technology described herein can be embodied in a method that includes receiving an audio signal encoding a portion of an utterance, and providing, to a first neural network, data corresponding to the audio signal. The method also includes generating, by a processor, data representing a transcription for the utterance based on an output of the first neural network. The first neural network is trained using features of multiple context-dependent states, the context-dependent states being derived from a plurality of context-independent states provided by a second neural network.
-
-
Search Results
3
records
(took 0.012 seconds)
