Unlock instant, AI-driven research and patent intelligence for your innovation.

Cross-lingual initialization of language models

A language model and language technology, applied in natural language translation, speech analysis, speech recognition, etc., can solve problems such as not having appropriate language model access

Active Publication Date: 2014-03-12
GOOGLE LLC
View PDF3 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

For somewhat infrequently used languages ​​or infrequently occurring contexts, the ASR engine may not have access to an appropriate language model

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Cross-lingual initialization of language models
  • Cross-lingual initialization of language models
  • Cross-lingual initialization of language models

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0018] figure 1 is a diagram of an example system 100 that machine-translates existing corpora or log recognition results associated with each corpus for estimating a language model of a target language and context. Briefly, the system 100 identifies an existing corpus that includes speech recognition results for a given language and target context. A target corpus can be generated by machine translating speech recognition results from an existing corpus from a given language into a different language, and optionally blending the machine-translated speech recognition results with other data sources in the target language. The target corpus can then be used to estimate language models specific to different languages ​​and the same target context.

[0019] System 100 includes client devices 108 and 110, such as cell phones, PDAs, e-book readers, smartphones, music players, or personal computers. Client devices 108 and 110 are configured to communicate with servers 116 , 118 ,...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for initializing language models for automatic speech recognition. In one aspect, a method includes receiving logged speech recognition results from an existing corpus that is specific to a given language and a target context, generating a target corpus by machine-translating the logged speech recognition results from the given language to a different, target language, and estimating a language model that is specific to the different, target language and the same, target context, using the target corpus.

Description

[0001] Cross References to Related Applications [0002] This application claims priority to US Application Serial No. 13 / 093,176, filed April 25, 2011, and entitled CROSS-LINGUAL INITIALIZATION OF LANGUAGE MODELS, the disclosure of which is incorporated herein by reference. Background technique [0003] An Automatic Speech Recognition (“ASR”) engine converts speech into text. In doing so, an ASR engine typically relies on an acoustic model that maps the sound of each utterance to candidate words or phrases, and specifies which of those candidate words or phrases are most likely to be correct based on the words or phrases' historical usage . [0004] To improve recognition accuracy, the ASR engine uses different acoustic models and language models to recognize utterances associated with different contexts. For example, one language model can be used to recognize utterances spoken by a user when entering a text message, while a different language model can be used when a use...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G10L15/183
CPCG10L15/197G06F17/289G10L15/005G10L15/183G06F40/58
Inventor 中嶋海介B·斯特罗普
Owner GOOGLE LLC