Scientific research in the field of the Internet content analysis is extremely relevant at present. This is due to the increasing and solidifying trend of human interaction with virtual reality and the user's involvement in the virtual environment. Especially important today is the development of socio-cyberphysical systems in terms of analysis of heterogeneous Internet content: text, images, video. This article presents the results of the study of the methods for analyzing the text of an Internet portal in order to identify the destructive impact on the user. In the course of the work, a parser was developed that allows one to obtain an actual text corpus (datasets) in Russian, consisting of users' comments to posts. Text markup and pre-processing were carried out through a specially developed software module for manual markup. The final part of the article presents the results of experiments to assess the accuracy of methods for content classifying to identify the facts of destructive character.