Development of Methods for Sentiment Analysis of Russian-language Texts

This work is devoted to study of machine learning state-of-the-art models applied to sentiments analysis of russian-language texts problem. The main tasks of this research are to analyze the efficiency of last developed neural network models solving the problem mentioned above, to compare obtained results with previous researches conducted with the use of public datasets and combination of datasets in order to get an universal sentiment analysis system for russian language. We use public russian-language sentiment analysis datasets to test our models. This work consists of analytical, theoretical and practical parts. To begin with, some recent language models are considered. After that, methods for distributed neural models learning are described. Finally, the comparison of results of this work and previous models applied to public datasets and to joined dataset are shown.