This article discusses the application of the frequency-context classification algorithm to texts of various styles. The main features of different styles that affect the efficiency of the algo-rithm are highlighted. It is proved that the method of selecting the subject of the text using the fre-quency-context classification algorithm works best in relation to scientific and legal documents and, in its current form, is practically inapplicable for literary texts. This makes the task of modifying the algorithm to determine the subject of literary texts relevant.

The article considers a mathematical model of a system that provides recognition of images that represent text or use similar information in the generation process.