Improving the Quality of Scientific Articles Machine Translation While Writing Original Text

摘要

Currently, the publication of scientific articles in peer-reviewed journals in most cases assumes translation into English. In this context, it is possible to reduce translation costs by using machine translation systems with post-editing. Modern machine translation systems have come a long way from bilingual dictionaries to multi-lingual localization platforms and continue to develop rapidly. By the use of neural networks deep learning, translation quality has almost reached the level of manual translation. However, machine translation has disadvantages and limitations, understanding of which allows using such systems to reduce time and costs for scientific texts translation. The paper formulates and systematizes main features of machine translation application, as well as gives recommendations for preparing a source text for translation, which are relevant not only when using machine translation systems, but also for human translation, especially for texts of specific subjects. By the example of a scientific and technical text and performing an experiment it is demonstrated how the quality of machine translation changes when described recommendations are applied. BLUE score metric is used to evaluate translation quality.