The Improvement of Machine Translation Quality with Help of Structural Analysis and Formal Methods-Based Text Processing

Abstract

This article considers the issues of enhancing the quality of machine translation from one language into another one by structuring linguistic patterns and using dentification methods for the situations that cannot be processed by the suggested approach and are subject to individual processing. According to the BLEU score metrics, the described approach allows to increase the quality of machine translation on average by 0.1 and reduce postprocessing time due to the identification of idioms and words with context-dependent meanings by translation. The experiment data base of the study was built upon online available pairs of texts that cover the events of FIFA World Cup 2018 and well-known idioms.