Automatic Classification of Russian-Language Texts from the Internet by Genre

This article is about using modern language models based on BERT and on three types of textual linguistic attributes for the automatically determining text genres as well presents a comparative analysis of these models from the standpoints of computer and classical linguistics. A corpus of Russian-language texts from the internet in eight genres has been collected, including postings on vk.com, comments and articles from habr.com, company descriptions, news, scientific articles, advertisements, and movie reviews from the Kinopoisk website. Each text is represented as a vector of numerical features, using each of the following selected models: five BERT variations and the linguistic features of symbolic, structural, and rhythmic levels.