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

Automatic translation quality evaluation method fusing syntactic information

A translation quality and automatic evaluation technology, applied in the field of translation, can solve the problems of restricting the effect of the model and grammatical features are rarely taken into account

Pending Publication Date: 2021-10-19
XIAMEN UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, grammatical features are rarely taken into account in translation quality estimation, restricting the effect of the model

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
  • Automatic translation quality evaluation method fusing syntactic information
  • Automatic translation quality evaluation method fusing syntactic information
  • Automatic translation quality evaluation method fusing syntactic information

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0026] The following embodiments will further illustrate the present invention in conjunction with the accompanying drawings.

[0027] The embodiment of the present invention is divided into three steps: obtaining word vector representation; syntactic graph generation and graph neural network; quality prediction model. At the same time, these three steps are also reflected in the figure 1 In the model architecture diagram of , the three parts are described below.

[0028] Word vector representation:

[0029] First, the input bilingual text representation vector needs to be obtained. This invention proposes to use a bilingual pre-training model, such as XLM-R (Conneau et al., 2019) or mBERT (Devlin et al., 2018) to obtain the word vector representation of the input text, and fine-tune the parameters during the model training process. Of course, the method of using Word2Vec is also feasible. You can use the open source toolkit Transformers (reference: https: / / github.com / hugg...

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 an automatic translation quality evaluation method fusing syntactic information, and relates to the technical field of translation. The method comprises the following steps: acquiring a bilingual text expression of an input text; constructing syntactic dependency trees for bilingual input texts to form syntactic graphs; using a graph neural network to encode related node relation features and then splicing the related node relation features, and connecting a simple sigmoid layer on the upper layer to output a quality score; and using the output of the model and the root-mean-square error of the data label as loss, and updating quality prediction model parameters through a back propagation algorithm. A graph neural network is utilized to ingeniously solve the problem of lack of introduction of syntactic information in automatic evaluation of translation quality, and the method is not seen in the field of automatic evaluation of translation quality. On the basis of the pre-training model, graph neural network coding syntactic information is added, so that the model can express semantic and syntactic information at the same time, and the effect of generally improving the Pearson's correlation coefficient by about 19% can be achieved compared with the effect of independently using the pre-training model.

Description

technical field [0001] The invention relates to the technical field of translation, in particular to an automatic translation quality evaluation method that integrates syntactic information. Background technique [0002] With the development of neural machine translation and natural language technology, how to automatically quantify translation quality (translation quality estimation, QE) has attracted widespread attention from the business community and academia. Automatic evaluation of translations based on big data-driven estimates of translation quality without reference to translations. At present, the methods of QE can be mainly classified into three categories: ① based on feature engineering; Representative, the disadvantage is that the performance is limited and it is difficult to deal with new language phenomena; method ② is usually two-stage, training a bilingual model on the basis of a large number of parallel corpora to obtain word expressions, and then inputtin...

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): G06F40/58G06F40/211G06F40/30G06N3/04G06N3/08
CPCG06F40/58G06F40/211G06F40/30G06N3/084G06N3/048G06N3/045
Inventor 陆晓蕾倪斌韩潮张培欣管新潮李力陈晨
Owner XIAMEN UNIV