Method for extracting single-word translation correspondences from parallel texts using distributional semantics models

The paper deals with problems of corpus research of linguistic units. The task of extracting translation correspondences from a parallel corpus is defined. An overview of existing approaches to this task is provided. The paper focuses on the approach to extracting translation correspondences based on distributional semantics models. The paper describes the theoretical model developed by the authors as well as its software implementation. A test parallel corpus of patent texts in French and English was compiled for the purpose of this research. The paper provides results of an experiment aimed at extracting single-word translation correspondences from the test parallel corpus.