Аннотация:This article examines the work of two Python libraries TextBlob and Dostoevsky for determining the sentiment of reviews in Russian. The TextBlob library does not directly provide sentiment analysis of texts in Russian, but it has the ability to transfer such texts to the input of the Google Translate translator, in order to then evaluate the polarity of these texts based on the generated English translations. The second Dostoevsky library was created specifically for sentiment analysis of texts in Russian and is trained on the largest Russian-language sentiment corpus RuSentiment, which contains more than 30 thousand manually annotated messages. To compare these libraries, this article uses real feedback from residents of the city of Nur-Sultan about the quality of life in the capital, including about the activities of city services and the housing and communal sector, the work of service facilities, education, health care, culture, etc., posted on public Facebook groups. The results of the sentiment analysis carried out are evaluated using generally accepted metrics of accuracy, completeness and F-measure and demonstrate a slight superiority of the Dostoevsky library.
