Corpus for text classification for domain of knowledge and style of writing.

Recommendation systems (RS) are, of course, the most commonly used to enhance various activities. These systems assist users by offering personalized recommendations based on their interests and requirements. We have developed a system that generates content using web scraping and automatic text summarization. In order to select the most valuable text, we need appropriate metrics based on text analysis. Thus, this article proposes a system of indicators to recommend texts for further summarization. It begins with a classification of recommendation systems and a general review of the content generation system.The <b>purpose</b> of this study is to develop a recommendation system for content generation, as well as to test the first module of the system – “Context of the Sources”. The <b>object</b> <b>of analysis </b>is a set of algorithms, services, or other software products used to determine a particular user's preferences. The <b>subject of the study</b> is natural language processing (NLP) methods in conjunction with the methods of supervised learning—classification.The result of the study is an empirical assessment of the first “Context of the Sources” module of RS, covering academic, security, and non-security domains using a text classification approach. In conclusion, ideas on the results of the experiment and the prospects for implementation of RS are formulated.