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

Modification method and device of attention parameter in neural machine translation, and electronic equipment

A machine translation and attention technology, applied in the software field, can solve problems such as over-translation and under-translation

Active Publication Date: 2018-11-02
BEIJING SOGOU TECHNOLOGY DEVELOPMENT CO LTD
View PDF4 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] Embodiments of the present invention provide a method, device, and electronic device for correcting attention parameters in neural machine translation, which are used to solve the technical problems of "over-translation" and "under-translation" in neural machine translation in the prior art, and improve neural machine translation. The accuracy of machine translation

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
  • Modification method and device of attention parameter in neural machine translation, and electronic equipment
  • Modification method and device of attention parameter in neural machine translation, and electronic equipment
  • Modification method and device of attention parameter in neural machine translation, and electronic equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0081] Please refer to figure 1 , the embodiment of the present application provides a method for correcting attention parameters in neural machine translation, the method comprising:

[0082] S11: Obtain the encoding vector of the source word to be translated at the tth moment;

[0083] S12: Obtain the hidden state vector of the target word translated at the t-1th moment, the coverage rate of the source word at the t-1th moment, and the relationship between the source word and the target word word alignment probability;

[0084] S13: Based on the word alignment probability, correct the coverage rate to obtain the corrected coverage rate;

[0085] S14: Based on the corrected coverage rate, the encoding vector and the hidden state vector, calculate and obtain attention parameters in the process of neural machine translation;

[0086] S15: Perform machine translation based on the attention parameter.

[0087] In the specific implementation process, neural machine translation...

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 modification method and device of an attention parameter in neural machine translation, and electronic equipment. The method comprises the following steps: acquiring a codingvector of a to-be-translated source-end word at the t-th moment, wherein t is the integer greater than or equal to 1; acquiring implicit state vector of the target end word acquired by translating atthe t-1-th moment, the coverage rate of the source end word at the t-1 moment, and the word alignment probability between the source end word and the target end word; based on the word alignment probability, modifying the coverage rate to obtain the modified coverage rate; computing to acquire the attention parameter in the neural machine translation based on the modified coverage rate, the coding vector and the implicit state vector; and performing neural machine translation based on the attention parameter. In the above technical scheme disclosed by the invention, the coverage rate in the neural machine translation is modified through the word alignment probability, thereby improving the accuracy of the attention parameter, the technical problem that over-translation and under-translation are easy to occur in the neural machine translation in the prior artis solved, and the translation accuracy is improved.

Description

technical field [0001] The invention relates to the field of software technology, in particular to a method, device and electronic equipment for correcting attention parameters in neural machine translation. Background technique [0002] With the development of economic globalization, the demand for cross-language translation is increasing, and the requirements for the accuracy of machine translation are getting higher and higher. Nowadays, machine translation with higher accuracy mainly includes Statistical Machine Translation (SMT) and Neural Machine Translation (NMT). [0003] Statistical machine translation is to build a statistical translation model through statistical analysis of a large number of parallel corpora, and then use this model for translation. Neural machine translation is to vectorize sentences in one language, convert them into a representation form that computers can "understand", and then decode them into translations in another language through conduc...

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): G06F17/28G06N3/08
CPCG06N3/08G06F40/45G06F40/58
Inventor 程善伯王宇光姜里羊陈伟王砚峰
Owner BEIJING SOGOU TECHNOLOGY DEVELOPMENT CO 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