公开(公告)号：US20180122367A1

公开(公告)日：2018-05-03

申请号：US15336056

申请日：2016-10-27

Applicant: Google Inc.

Inventor： Carl-Anton Ingmarsson

IPC: G10L15/187 ,  G10L25/15 ,  G10L15/01 ,  G10L15/18 ,  G10L15/30

CPC classification number: G10L15/187 ,  G10L15/01 ,  G10L15/065 ,  G10L15/1815 ,  G10L15/183 ,  G10L15/30 ,  G10L25/15

Abstract: In some implementations, systems and methods are capable of performing location-based voice query recognition. Audio corresponding to an utterance of the user and data indicating a location associated with a user are initially received. An initial transcription of the utterance is then obtained. A set of n-grams that are associated with a region are selected for each of multiple regions that are associated with the location. One or more additional candidate descriptions are generated for one or more of the multiple geographic regions. A representative transcription of the utterance is selected for one or more of the multiple geographic regions. The representative transcription is finally provided to an automated speech recognizer.

公开(公告)号：US09576578B1

公开(公告)日：2017-02-21

申请号：US14824902

申请日：2015-08-12

Applicant: Google Inc.

Inventor： Gleb Skobeltsyn ,  Alexandru Ovidiu Dovlecel ,  Carl-Anton Ingmarsson ,  Martin Baeuml ,  Behshad Behzadi ,  Dmitry Osmakov

IPC: G10L15/26 ,  G10L15/08 ,  G06F17/30

CPC classification number: G10L15/26 ,  G06F17/30684 ,  G06F17/30746 ,  G10L15/22 ,  G10L2015/226

Abstract: Methods, including computer programs encoded on a computer storage medium, for collaborative language model biasing. In one aspect, a method includes: obtaining (i) one or more initial candidate transcriptions, and (ii) one or more terms that are associated with a context; selecting one or more of the terms that are associated with the context, and that (i) do not occur in the candidate transcriptions, and (ii) are indicated as phonetically similar to one or more terms that do occur in the initial candidate transcriptions; generating one or more additional candidate transcriptions based on the (i) initial candidate transcriptions, and (ii) the selected terms; and providing the one or more additional candidate transcriptions to an automated speech recognizer.

Abstract translation: 方法，包括在计算机存储介质上编码的计算机程序，用于协作语言模型偏移。 一方面，一种方法包括：获得（i）一个或多个初始候选转录，以及（ii）与上下文相关联的一个或多个术语; 选择与上下文相关联的一个或多个术语，并且（i）不在候选转录中发生，并且（ii）被表示为与在初始候选转录中确实发生的一个或多个术语的语音相似; 基于（i）初始候选转录，和（ii）所选择的术语生成一个或多个另外的候选转录; 以及将一个或多个附加候选转录提供给自动语音识别器。