English translation Chinese word sense disambiguation method based on neural network

A neural network and word meaning disambiguation technology, applied in the field of machine translation, can solve problems such as unsatisfactory results and no obvious boundaries of attention, and achieve the effect of improving English reading efficiency

Inactive Publication Date: 2019-09-27
吕海港
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, since this attention mechanism is a soft alignment, the attention between word senses has no clear boundary
The meanings found by this method are often ...

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
  • English translation Chinese word sense disambiguation method based on neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0015] The embodiment of the present invention adopts the open source neural network machine translation software OpenNMT software package (http: / / opennmt.net / ), and the 1 million Chinese-English corpus used for training comes from the open source Niutrans software package (http: / / www.niutrans.com ), the English-Chinese dictionary comes from the ECDict project (https: / / github.com / skywind3000 / ECDICT).

[0016] This embodiment mainly includes two parts ( figure 1 ): training translation models and decoding word sense disambiguation.

[0017] In the stage of training the translation model, it is divided into four steps.

[0018] The first step is to extract English words and their variants according to ECDict’s English-Chinese dictionary file, find out all the corresponding Chinese meanings, and generate an English-Chinese dictionary, one word per line, for example, the format of all meanings of the word work and its variants is as follows :

[0019] work|||work, work, labor, ...

Embodiment 2

[0030] The English-Chinese dictionary, corpus processing, and decoding process are the same as in Example 1. The training translation model is trained using a two-layer convolutional neural network (CNN) with 500 hidden units and a global attention mechanism. Both the source language and the target language use 100,000 words amount, each layer uses a 512-dimensional word vector space, and the generated translation model is about 900MB. Using this translation model for restricted decoding, the accuracy of word meaning in sentences is 79.4%.

Embodiment 3

[0032] The English-Chinese dictionary, corpus processing, and decoding process are the same as those in Example 1. The training translation model uses a 6-layer 512-hidden unit transformer (Transformer) and 8 multi-head self-attention mechanisms for training. Both the source language and the target language use 100,000 The amount of words, each layer uses a 512-dimensional word vector space, and the generated translation model is about 3000MB. Using this translation model for restricted decoding, the accuracy of word meaning in sentences is 83.2%.

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

In order to determine the accurate Chinese word sense of words in English sentences, the invention provides an English translation Chinese word sense disambiguation method based on neural network, which comprises the following steps of: firstly, generating a Chinese and English mixed sequence corresponding to each English sentence based on an English-Chinese dictionary and Chinese and English sentences of Chinese and English corpora; secondly, taking an English sentence and Chinese and English mixed sequence as parallel corpora, and performing training through a neural network method to obtain a translation model; and finally, sequentially performing restrictive decoding in the Chinese meaning of each English word of the to-be-translated English sentence by using the translation model to find the accurate meaning of the word in the English sentence, thereby effectively solving the problem of word meaning ambiguity elimination in English translation.

Description

technical field [0001] The invention relates to the field of machine translation, in particular to a word sense disambiguation method for English-Chinese translation. Background technique [0002] Word sense disambiguation is an important and difficult point in natural language processing. In English-Chinese translation, an English word can have one or more Chinese meanings, and finding the exact meaning of the word in the current sentence is an unsolved problem. [0003] At present, neural network-based machine translation is relatively mature. It can translate English sentences into Chinese sentences relatively accurately, and then use the attention mechanism to align English sentences and Chinese sentences at the word level, so as to find the approximate meaning of words. However, since this attention mechanism is a kind of soft alignment, there is no clear boundary of attention between word senses. The meaning of words found by this method is often one or two character...

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): G06F17/27G06F17/28G06N3/04G06N3/08
CPCG06N3/08G06F40/30G06F40/58G06N3/044G06N3/045
Inventor 吕海港
Owner 吕海港
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