АКТУАЛЬНЫЕ ПОДХОДЫ К УЛУЧШЕНИЮ ТОЧНОСТИ НЕЙРОННОГО МАШИННОГО ПЕРЕВОДА

This article reviews topical approaches dealing with the problem of accuracy of neural machine translation. Classification of these approaches is provided in two groups: structural and linguistic. Moreover, the article sheds light on the most common problems of systems with neural machine translation. Effectiveness of the methods reviewed in this article is assessed based on BLEU metrics for the English-German language pair. Conclusions about the assessment of effectiveness of these methods are given in the closing part of the article.