N. V. Loukachevitch’s research while affiliated with Lomonosov Moscow State University and other places

The results of applying standard neural network architectures (CNN, LSTM, and BiLSTM) and the recent BERT models to previously labeled data for sentiment analysis in Russian are described. Two versions of the Russian-language BERT model (RuBERT) and different methods of its application to the problem of sentiment analysis are compared. The state-of-the-art results on the datasets were improved significantly.

Описываются результаты применения стандартных архитектур нейронных сетей (CNN, LSTM, BiLSTM) и недавно появившихся их моделей BERT на ранее размеченных данных для анализа тональности текстов на русском языке. Сравниваются два варианта русскоязычной модели BERT (RuBERT) и различные способы ее применения.

Texts can convey several types of inter-related information concerning opinions and attitudes. Such information includes the author’s attitude towards mentioned entities, attitudes of the entities towards each other, positive and negative effects on the entities in the described situations. In this paper, we described the lexicon RuSentiFrames for Russian, where predicate words and expressions are collected and linked to so-called sentiment frames conveying several types of presupposed information on attitudes and effects. We applied the created frames in the task of extracting attitudes from a large news collection.