Resumen
The paper deals with the application of distributional semantic methods to the task of measuring similarity between several translations of the original text. In particular, Word2Vec neural network toolkit is employed for comparison between two translations. Moreover, in terms of the theory of translation, descriptions of transformations for paraphrasing, which are also used for testing plagiarism detection methods, suit the task of comparing translations. Experiments discussed in this paper are carried out for the Russian original and English translations of M. Bulgakov's novel "The Master and Margarita". In the paper, the above mentioned approaches are combined to contrast the translation by M. Glenny (1967) with one by R. Pevear and L. Volokhonsky (1997). Hypothesis that parallel translations can be treated as paraphrases obtained as a result of transformations is under consideration. The paper contains detailed quantitative analysis of the data obtained regarding the similarity between two translations of fiction text as well as discussion of particular contexts.