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

A method for calculating the accuracy of speech recognition based on triplets

A recognition accuracy and speech recognition technology, applied in the field of data processing, can solve problems such as errors in recognition results, and achieve the effect of reducing deviation and enriching semantic information

Active Publication Date: 2021-04-20
北京海天瑞声科技股份有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] Obviously, the above two strings represent the same year, month, and day, but when WER is calculated, the recognition result is considered to be wrong

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
  • A method for calculating the accuracy of speech recognition based on triplets
  • A method for calculating the accuracy of speech recognition based on triplets
  • A method for calculating the accuracy of speech recognition based on triplets

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0049]The exemplary embodiment will be described in detail herein, and examples thereof are shown in the drawings. The following description is related to the drawings, unless otherwise indicated, the same numbers in the drawings represent the same or similar elements. The embodiment described in the exemplary embodiments will not represent all embodiments consistent with the present invention. In contrast, examples of apparatus and methods consistent with some aspects of the invention, as detailed in the appended claims.

[0050]At present, in some technologies, the current domestic and foreign use of the accuracy calculation method for speech recognition is the word error rate WER (Word ERROR RATE). The calculation method of WER is as follows: Test data contains "correct answer" (text) and corresponding voice, input voice input to the identification engine, obtain "identification result", then calculate "identification results" relative to "correct answer" in the text level Wer.

[0051...

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 present disclosure relates to a triplet-based speech recognition accuracy calculation method and device, electronic equipment, and a computer-readable storage medium. Among them, the method includes: obtaining the speech to be recognized and the correct text corresponding to the speech; recognizing the speech to obtain the recognized text; labeling the recognized text and the correct text; analyzing the marked recognized text and the correct text, and obtaining the triple sequence of the recognized text and the correct text The triplet sequence; calculate the recognition accuracy rate based on the triplet sequence of the recognized text and the triplet sequence of the correct text. On the one hand, marking the text can effectively divide the words that will affect the accuracy of speech recognition; on the other hand, the speech recognition results and correct answers are unified into triplets, which further enriches the semantic information of words and reduces the need for recognition The different formats of engine recognition results lead to statistical deviations in the accuracy rate, which provides a quality control basis for the improvement of the speech recognition engine.

Description

Technical field[0001]The present disclosure relates to the field of data processing, and more particularly to a speech recognition accuracy calculation method and apparatus, an electronic device, and a computer readable storage medium based on a terpolyzal speech recognition.Background technique[0002]At present, with the continuous development of speech recognition technology, speech-recognized applications are increasingly wide, and have gradually penetrated into family life, office fields, entertainment and other aspects. The user enters voice (for example, reading one sentence) by using the microphone of the individual computer, laptop, tablet, dedicated learning terminal, smartphone, etc., and via the voice recognition engine (hereinafter referred to as "engine" ) Complete voice - text conversion. For the speech recognition engine, the public is generally concerned about its identification accuracy, that is, whether the identified text is consistent with the correct text of the ...

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 Patents(China)
IPC IPC(8): G10L15/26G10L25/69G06F16/35G06F40/205G06F40/279G06F40/30G06N3/04
CPCG10L15/26G10L25/69G06F16/35G06F40/205G06F40/279G06F40/30G06N3/044G06N3/045
Inventor 杜沪邵志明宋琼李科郝玉峰黄宇凯
Owner 北京海天瑞声科技股份有限公司