Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Speech recognition method, device and apparatus

A speech recognition and speech technology, applied in speech recognition, speech analysis, instruments, etc., can solve problems that are difficult to include language phenomena and affect the accuracy of speech recognition, and achieve the effect of improving recognition accuracy and saving the training process

Pending Publication Date: 2019-03-26
ALIBABA GRP HLDG LTD
View PDF6 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] The quality of language model training has an important impact on the performance of speech recognition. The larger the training corpus, the better the effect of speech recognition, but no matter how large the training corpus is, it is difficult to include all language phenomena.
Although some fields can improve the accuracy of speech recognition in this field by training the language model on the corpus in the field, for some specific words, especially the appearance of hot words and the time period of hot words (some words in a certain period of time) Some events are often mentioned by people, such as the title of a new song), which still greatly affects the accuracy of speech recognition

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
  • Speech recognition method, device and apparatus
  • Speech recognition method, device and apparatus
  • Speech recognition method, device and apparatus

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0049] A method of speech recognition, such as figure 1 shown, may include:

[0050] Step 101, receiving voice from the user;

[0051] Step 102, acquiring a hot word language model, the hot word language model is a language model trained according to hot words provided by users;

[0052] Step 103, using the hot word language model and the preset main language model to decode the speech.

[0053]In the method of this embodiment, the hot word language model is compiled through word segmentation and vocabulary, and then the hot word language model is combined with the existing main language model for decoding. The superposition of language model scores greatly improves the recognition accuracy of hot words without affecting the recognition rate of the overall word sequence, solves the problem of low recognition rate and poor recognition effect of hot words, and can immediately and quickly respond to any occurrences in various application scenarios hot words; in addition, the t...

Embodiment 2

[0085] This embodiment provides a voice recognition device, such as Figure 4 As shown, can include:

[0086] Receiving module 41, is used for receiving the speech from user;

[0087] Obtaining module 42, is used for obtaining hot word language model, and described hot word language model is the language model that obtains according to the hot word training that user provides;

[0088] The decoding module 43 is configured to use the hot word language model and the preset main language model to decode the speech.

[0089] In this embodiment, the acquisition module 42 may acquire hot word language models in various ways. In one implementation manner, the acquisition module 42 may be configured to obtain a vocabulary of hot words according to hot words and weight information provided by users, and compile a language model of hot words according to the vocabulary of hot words. In another implementation, the acquisition module 42 can be used to obtain the hot word vocabulary acc...

Embodiment 3

[0095] A speech recognition device, comprising:

[0096] Stored with a speech recognition program memory;

[0097] A processor configured to read the speech recognition program to perform the following operations:

[0098] receive voice from the user;

[0099] Acquiring a hot word language model, the hot word language model is a language model trained according to the hot words provided by the user;

[0100] The voice is decoded by using the hot word language model and the preset main language model.

[0101] The speech recognition device in this embodiment may be any computing device capable of realizing the above functions. In practical applications, the computing device may be a physical server, a virtual server, a distributed system formed by a physical server or a virtual server, and the like.

[0102] For other details of this embodiment, refer to Embodiment 1.

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

The invention discloses a speech recognition method, device and apparatus. The method comprises the following steps: receiving voice from a user; obtaining a hot word language model, wherein the hot word language model is a language model obtained by training according to hot words provided by the user; decoding the voice by using the hot word language model and a preset main language model. According to the method, at least the hot word recognition accuracy can be effectively improved.

Description

technical field [0001] The present invention relates to the technical field of speech, in particular to a speech recognition method, device and equipment. Background technique [0002] The quality of language model training has an important impact on the performance of speech recognition. The larger the training corpus, the better the effect of speech recognition, but no matter how large the training corpus is, it is difficult to cover all language phenomena. Although some fields can improve the accuracy of speech recognition in this field by training the language model on the corpus in the field, for some specific words, especially the appearance of hot words and the time period of hot words (some words in a certain period of time) Some events are mentioned more by people, such as the title of a new song), which still greatly affects the accuracy of speech recognition. Contents of the invention [0003] The present application aims to solve at least one of the technical ...

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
IPC IPC(8): G10L15/06G10L15/30G10L15/32
CPCG10L15/063G10L2015/0635G10L15/30G10L15/32Y02D10/00
Inventor 高杰李威朱林
Owner ALIBABA GRP HLDG LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com