Speech translation method and system

A speech translation and speech technology, applied in the field of speech translation methods and systems, can solve the problems of inability to obtain enough multi-source languages, inability to meet real-time requirements, limited practicability, etc., to increase readability, reduce training difficulty, The effect of reducing the scale

Active Publication Date: 2020-10-16
XIAMEN KUAISHANGTONG TECH CORP LTD
View PDF5 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] (1) The process is complex and requires a lot of preparatory work;
[0004] (2) It cannot meet the needs of scenarios with high real-time requirements;
[0005] (3) Because of the cascade structure, the error in the first stage will propagate to the second stage, affecting the final effect;
[0006] (4) In many cases, the source language is a small language, and it is impossible to obtain enough texts in the source language for training speech recognition models and machine translation models, and the practicability is limited

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 translation method and system
  • Speech translation method and system
  • Speech translation method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0055] This embodiment provides a speech translation method for transcribing Hokkien speech into target language text.

[0056] The speech translation model of the present invention adopts a sequence-to-sequence architecture based on Transformer, as attached figure 1 shown.

[0057] In the encoder part on the left, the input speech features are converted into another dimensional vector (InputEmbedding) by transformation, and summed with the positional encoding (Positional Encoding) vector, and then through a series of matrix operations in the dotted line box, more Abstract features, as the output of the encoding part, the operation in the dotted box can be repeated multiple times, and then the output of the encoding part is used as part of the input of the multi-head attention layer (Multi-Head Attention) of the decoder part on the right.

[0058] In the decoder part on the right, the recognized text of the previous time step is used as input, first transformed into ...

Embodiment 2

[0099] This embodiment provides a speech translation system, including:

[0100] Speech input terminal, the user inputs speech to be recognized in a small language through components such as a microphone, the terminal extracts the speech features of the speech to be recognized, and transmits the speech features to the recognition module.

[0101] The recognition module stores a list of replaceable words and a dictionary of subwords, and is loaded with a speech translation model model, calculates the position encoding vectors for the above speech features and sums them up, transcribes the speech of the minority language into the target language text, and outputs the recognized text.

[0102] Applying this system to APPs or other smart devices on mobile phones can meet the needs of users for voice translation in small languages.

[0103] Those skilled in the art can understand that all or part of the steps in the embodiment of the above voice data detection method can ...

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 translation method and a speech translation system. The speech translation method comprises the following steps of: constructing a small language and mandarin replaceable word list; constructing a sub-word dictionary; carrying out one-hot coding on each character; to-be-recognized voice is acquired, and voice features are extracted; calculating a position coding vector of the to-be-recognized voice feature; summing the to-be-recognized speech features and the position coding vectors; inputting a to-be-recognized input vector into the trained speech translationmodel; and outputting the recognition text by the speech translation model. According to the invention, the training difficulty of the speech translation model is reduced and the training speed is improved.

Description

technical field [0001] The invention relates to the technical field of voice recognition, in particular to a voice translation method and system. Background technique [0002] In many cases, it is necessary to transcribe the voice of a small language that only a few people can understand into a text that most people can understand, such as transcribing the Hokkien voice into Mandarin. This technology of transcribing the speech of the source language into the text of the target language usually includes two cascading processes: first, the speech of the source language is transcribed into text of the source language through language recognition technology, and then the speech of the source language is translated into text of the source language by machine translation technology. The text is translated into text in the target language. But this two-stage cascade system has the following problems: [0003] (1) The process is complex and requires a lot of preparatory work; [...

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): G06F40/42G10L15/02G10L15/08G10L15/26
CPCG06F40/42G10L15/08G10L15/02Y02T10/40
Inventor 徐敏肖龙源李稀敏蔡振华刘晓葳谭玉坤
Owner XIAMEN KUAISHANGTONG TECH CORP LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap