Resumen
Topic modeling is one of the most widely used methods in text analysis. It can be used to select topics as well as to find the topics distributed in each document from the corpus. In this article, we present a method for clustering communities in the social network VKontakte (the most popular Russian social network) using topic modeling. As a communities sample a set of groups for which several students of Tomsk State University are subscribed was selected. There were about 7,000 of them in this set. The article describes the method by which the text corpus was formed, as well as mathematical modeling using two popular classical methods LDA and ARTM. A detailed description of these models, quality assessment criteria, and the main practical techniques used by the authors in training the models are given. The aggregated results of clustering communities by topic are also presented. There are also described a method for expert evaluation of community topics based on visualization of the words that make up the lexical core of the topic.