Neural technologies of comparative translation in teaching students of oil and gas university on the basis of Russian and English languages

Abstract

In the context of globalization and intensification of scientific and technical cooperation between the countries, the issue of the need for high-quality education of specialists in the oil and gas industry is being actualized. According to the St. Petersburg State Mining University, about 68% of research in this field is published in English. This article examines the possibility of using neural technologies of comparative translation to improve the level of language training of students of oil and gas universities based on Russian and English. The developed system based on multilayer perceptrons and long-term short-Term Memory (LSTM) has shown efficiency in determining semantic relations between terms in oil and gas texts, achieving accuracy of up to 94.7% on a test sample of 800 documents. These data provide significant potential for the creation of new teaching methods, increasing the level of interdisciplinary competence of future specialists. Neural comparative translation technologies are a type of machine learning models that help solve problems of interlanguage translation. These models are pre-trained and use parallel data for pre-training, which gives them the ability to predict masked words on the source and target side with reference to the context of either side of the parallel data.