XML document translation and evaluation method based on attention mechanism

Abstract

The invention discloses an XML document translation and evaluation method based on an attention mechanism. The method comprises the following steps of adopting a translation model based on a self-attention mechanism for model training; constructing a parallel corpus based on an XML language, constructing a word list, and converting the obtained corresponding words from a one-hot vector to a word embedding vector; constructing an XML document preprocessing system, judging and setting whether sentences under various label types need to be translated or not, and replacing labels in the XML sentences needing to be translated with unified symbols; constructing an XML document post-processing system, and restoring uniform symbols in sentences generated by the translation model into labels corresponding to the input files; creating an XML document translation quality evaluation system and determining the accuracy of the label content and number of sentences translated by the model, the sentence translation quality and whether the positions of labels in the sentences are correct or not. Whether sentences under each label type are translated or not is set and selected, and the XML labels are used for constructing the user dictionary, so that the label translation problem is solved.

Description

Technical field[0001]The invention relates to a neural machine translation technology, in particular to an XML document translation and evaluation method based on an attention mechanism.Background technique[0002]Machine translation refers to the process of using a computer to convert one natural language into another natural language. With the development of technology, neural machine translation technology based on deep learning has become the mainstream. Compared with traditional statistical-based machine translation methods, neural machine translation translations perform better in terms of sentence fluency and sentence meaning accuracy. Judging from the BLEU value of the machine translation quality evaluation index, in the case of abundant corpus, the effect of neural machine translation is far better than statistical machine translation. And in some tasks, neural machine translation has achieved results comparable to professional translators.[0003]The achievements of neural mac...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the deficiency that the neural machine model in the prior art cannot handle the relationship between tags and translations in XML well, the technical problem to be solved by the present invention is to provide an XML document translation and evaluation method based on the attention mechanism, using human Translation habits, data preprocessing, postprocessing and other means have improved the translation of XML documents, and completed a new evaluation index to help evaluate the translation effect of XML documents and help to correct the translation effect more accurately

Images