Sentiment Analysis of Russian-Language Texts Using Neural Network Models
Irina Ivanovna Prosvirkina, Mariya Golubeva · Scientific Research and Development Modern Communication Studies · 2025
This article focuses on determining the emotional sentiment of Russian-language texts using neural network models, particularly DeepSeek. In the context of digitalization, identifying markers that represent the tone of a statement (negative or positive) has become increasingly relevant for two main reasons: first, it saves researchers’ time, and second, it ensures impartiality by eliminating authorial interpretation. However, existing language models, primarily trained on English-language corpora, show limited accuracy when applied to Russian texts—especially in detecting positive sentiment. The challenge of identifying tonal markers is further complicated by the stylistic diversity of linguistic expressions conveying positive or negative emotionality in user discourse. Thus, testing DeepSeek’s performance in a Russian-language digital environment helps reveal typical distortions in its interpretation of evaluative context and outlines potential improvements for existing neural network models in analyzing Russian discourse.