A Data Augmented Machine Translation Method Based on Similar Words and Synonym Replacement

Abstract

The invention is the process of processing or conversion technology of natural language. A data enhancement method based on the replacement of similar words and synonyms has been disclosed.Tables and synonymous vocabulary; build similar words and synonymous vocabulary in the word vector obtained during the training process of large -scale language, and then replace similar words and synonyms in scarce small words;And the neural network training small -language neural network machine translation model with a attention mechanism.Expanded training data. The parameters of the neural network translation model can be well learned in sufficient data, and it can alleviate the problem of un logical words in the translation of neural machines, so that the translation quality of the translation model is improved.When the quality of the entire network on the development and collection is no longer significantly improved, the network parameters have been well learned.

Description

technical field [0001] The invention belongs to the technical field of natural language processing or conversion, and in particular relates to a data-enhanced machine translation method based on the replacement of similar words and synonyms. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: With the improvement of computer computing power and the application of big data, deep learning has been further applied, and Neural Machine Translation based on deep learning has attracted more and more attention. As a research hotspot of artificial intelligence, machine translation has very important scientific research value and practical value. In the field of NMT, one of the most commonly used translation models is the encoder-decoder model with attention-based. The main idea is to encode the sentence to be translated (hereinafter collectively referred to as "source sentence") into a vector representation through an en...

AI Technical Summary

Problems solved by technology

[0004] (1) Large-scale, high-quality bilingual parallel corpus is difficult to obtain, and the cost of constructing high-quality bilingual parallel corpus by human translation is relatively high

[0005] (2) Lack of large-scale, high-quality bilingual parallel corpus, resulting in insufficient training data and poor performance of the neural network machine translation model in small languages, making it difficult to build a high-performance machine translation system

However, as a small language, both translation talents and translation systems are very scarce.

Images