公开(公告)号：US09076445B1

公开(公告)日：2015-07-07

申请号：US13705228

申请日：2012-12-05

Applicant: Google Inc.

Inventor： Matthew I. Lloyd

IPC: G10L15/18

CPC classification number: G10L15/265 ,  G10L15/18 ,  G10L15/183 ,  G10L15/22 ,  G10L25/12 ,  G10L25/51 ,  G10L2015/223

Abstract: Methods, systems, and apparatuses, including computer programs encoded on a computer storage medium, for adjusting language models. In one aspect, a method includes accessing audio data. Information that indicates a first context is accessed, the first context being associated with the audio data. At least one term is accessed. Information that indicates a second context is accessed, the second context being associated with the term. A similarity score is determined that indicates a degree of similarity between the second context and the first context. A language model is adjusted based on the accessed term and the determined similarity score to generate an adjusted language model. Speech recognition is performed on the audio data using the adjusted language model to select one or more candidate transcriptions for a portion of the audio data.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的用于调整语言模型的计算机程序。 一方面，一种方法包括访问音频数据。 访问指示第一上下文的信息，第一上下文与音频数据相关联。 至少有一个术语被访问。 访问指示第二上下文的信息，第二上下文与该术语相关联。 确定表示第二上下文与第一上下文之间的相似程度的相似度得分。 基于被访问的术语和所确定的相似性得分来调整语言模型以产生经调整的语言模型。 使用经调整的语言模型对音频数据执行语音识别，以针对一部分音频数据选择一个或多个候选转录。

公开(公告)号：US08682659B2

公开(公告)日：2014-03-25

申请号：US13862170

申请日：2013-04-12

Applicant: Google Inc.

Inventor： Trausti T. Kristjansson ,  Matthew I. Lloyd

IPC: G10L21/02

CPC classification number: G10L21/0208 ,  G10L15/20

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for enhancing speech recognition accuracy. In one aspect, a method includes receiving an audio signal that corresponds to an utterance recorded by a mobile device, determining a geographic location associated with the mobile device, identifying a set of geotagged audio signals that correspond to environmental audio associated with the geographic location, weighting each geotagged audio signal of the set of geotagged audio signals based on metadata associated with the respective geotagged audio signal, and using the set of weighted geotagged audio signals to perform noise compensation on the audio signal that corresponds to the utterance.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于增强语音识别精度。 在一个方面，一种方法包括接收对应于由移动设备记录的话语的音频信号，确定与移动设备相关联的地理位置，识别与地理位置相关联的环境音频对应的一组地理标记音频信号， 基于与相应的地理标记音频信号相关联的元数据，对所述一组地理标记音频信号的每个地理标记音频信号进行加权，并且使用该组加权的地理标记音频信号对对应于话语的音频信号执行噪声补偿。

公开(公告)号：US20150269938A1

公开(公告)日：2015-09-24

申请号：US14735416

申请日：2015-06-10

Applicant: Google Inc.

Inventor： Matthew I. Lloyd

IPC: G10L15/26 ,  G10L25/12 ,  G10L15/22 ,  G10L25/51

CPC classification number: G10L15/265 ,  G10L15/18 ,  G10L15/183 ,  G10L15/22 ,  G10L25/12 ,  G10L25/51 ,  G10L2015/223

Abstract: Methods, systems, and apparatuses, including computer programs encoded on a computer storage medium, for adjusting language models. In one aspect, a method includes accessing audio data. Information that indicates a first context is accessed, the first context being associated with the audio data. At least one term is accessed. Information that indicates a second context is accessed, the second context being associated with the term. A similarity score is determined that indicates a degree of similarity between the second context and the first context. A language model is adjusted based on the accessed term and the determined similarity score to generate an adjusted language model. Speech recognition is performed on the audio data using the adjusted language model to select one or more candidate transcriptions for a portion of the audio data.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的用于调整语言模型的计算机程序。 一方面，一种方法包括访问音频数据。 访问指示第一上下文的信息，第一上下文与音频数据相关联。 至少有一个术语被访问。 访问指示第二上下文的信息，第二上下文与该术语相关联。 确定表示第二上下文与第一上下文之间的相似程度的相似度得分。 基于被访问的术语和所确定的相似性得分来调整语言模型以产生经调整的语言模型。 使用经调整的语言模型对音频数据执行语音识别，以针对一部分音频数据选择一个或多个候选转录。

公开(公告)号：US20130297313A1

公开(公告)日：2013-11-07

申请号：US13862219

申请日：2013-04-12

Applicant: GOOGLE INC.

Inventor： Matthew I. Lloyd ,  Trausti T. Kristjansson

IPC: G10L15/22

CPC classification number: G10L15/22 ,  G10L15/065 ,  G10L15/30

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for enhancing speech recognition accuracy. In one aspect, a method includes receiving an audio signal that corresponds to an utterance recorded by a mobile device, determining a geographic location associated with the mobile device, adapting one or more acoustic models for the geographic location, and performing speech recognition on the audio signal using the one or more acoustic models model that are adapted for the geographic location.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于增强语音识别精度。 在一个方面，一种方法包括接收对应于由移动设备记录的话语的音频信号，确定与移动设备相关联的地理位置，调整用于地理位置的一个或多个声学模型，以及对该音频执行语音识别 使用适合于地理位置的一个或多个声学模型模型的信号。

公开(公告)号：US08543645B1

公开(公告)日：2013-09-24

申请号：US13653195

申请日：2012-10-16

Applicant: Google Inc.

Inventor： Matthew I. Lloyd ,  Douglas H. Beeferman

IPC: G06F15/16

CPC classification number: G06F11/3688 ,  G06F11/3006 ,  G06F11/3051 ,  G10L15/01

Abstract: This disclosure generally relates to assigning and simultaneously running multiple client-side experiments on client devices. A file includes information regarding experiments that are available, including information regarding “layers,” which are logical, imaginary containers in which each experiment “resides.” Each experiment is associated with one layer. For each experiment, the file includes information regarding a location and size of the experiment within the layer. When the client device takes an action, a software module identifies a value of an identifier associated with the action. Each such identifier is associated with one or more of the layers. The software module can calculate, for each of the associated layers, a location within the layer based on the identifier value. The computer software module can identify, based on the information in the file, each experiment that overlaps with the calculated location within each layer and cause each identified experiment to be activated.

Abstract translation: 本公开通常涉及在客户端设备上分配和同时运行多个客户端实验。 一个文件包括有关可用实验的信息，包括有关“层”的信息，这些信息是每个实验所在的逻辑，虚拟容器。 每个实验都与一层相关联。 对于每个实验，该文件包括有关该层内实验的位置和大小的信息。 当客户端设备采取动作时，软件模块识别与该动作相关联的标识符的值。 每个这样的标识符与一个或多个层相关联。 软件模块可以针对每个相关联的层，基于标识符值来计算层内的位置。 计算机软件模块可以根据文件中的信息识别与每个层内的计算位置重叠的每个实验，并使每个识别的实验被激活。

公开(公告)号：US09542945B2

公开(公告)日：2017-01-10

申请号：US14735416

申请日：2015-06-10

Applicant: Google Inc.

Inventor： Matthew I. Lloyd

IPC: G10L15/065 ,  G10L15/26 ,  G10L15/183 ,  G10L15/18 ,  G10L15/22 ,  G10L25/12 ,  G10L25/51

CPC classification number: G10L15/265 ,  G10L15/18 ,  G10L15/183 ,  G10L15/22 ,  G10L25/12 ,  G10L25/51 ,  G10L2015/223

Abstract: Methods, systems, and apparatuses, including computer programs encoded on a computer storage medium, for adjusting language models. In one aspect, a method includes accessing audio data. Information that indicates a first context is accessed, the first context being associated with the audio data. At least one term is accessed. Information that indicates a second context is accessed, the second context being associated with the term. A similarity score is determined that indicates a degree of similarity between the second context and the first context. A language model is adjusted based on the accessed term and the determined similarity score to generate an adjusted language model. Speech recognition is performed on the audio data using the adjusted language model to select one or more candidate transcriptions for a portion of the audio data.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的用于调整语言模型的计算机程序。 一方面，一种方法包括访问音频数据。 访问指示第一上下文的信息，第一上下文与音频数据相关联。 至少有一个术语被访问。 访问指示第二上下文的信息，第二上下文与该术语相关联。 确定表示第二上下文与第一上下文之间的相似程度的相似度得分。 基于被访问的术语和所确定的相似性得分来调整语言模型以产生经调整的语言模型。 使用经调整的语言模型对音频数据执行语音识别，以针对一部分音频数据选择一个或多个候选转录。

公开(公告)号：US09418177B1

公开(公告)日：2016-08-16

申请号：US13958740

申请日：2013-08-05

Applicant: Google Inc.

Inventor： Matthew I. Lloyd ,  Johan Schalkwyk ,  Pankaj Risbood

IPC: G10L21/00 ,  G06F17/30 ,  G10L15/26

CPC classification number: G06F17/30976 ,  G10L15/197 ,  G10L15/265

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for processing spoken query terms. In one aspect, a method includes performing speech recognition on an audio signal to select two or more textual, candidate transcriptions that match a spoken query term, and to establish a speech recognition confidence value for each candidate transcription, obtaining a search history for a user who spoke the spoken query term, where the search history references one or more past search queries that have been submitted by the user, generating one or more n-grams from each candidate transcription, where each n-gram is a subsequence of n phonemes, syllables, letters, characters, words or terms from a respective candidate transcription, and determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries, and a weighting value that is based on the respective frequency.