Automatic Analysis of Literary Texts in Russian

Nowadays, the volume of textual information that a person faces, starting from school years, is large. Texts that require reading, understanding the structure of what is written can be specialized textbooks or literary works at school, as well as documents, laws, letters, or instructions for workers in various professions. In order to understand the essence of these texts, without reading them completely, you can familiarize yourself with a shorter and more comprehensive presentation, which basically contains facts. In this paper, one of the ways to quickly understand the structure of interactions in a literary work is realized - the construction of a graph of relations between the main characters. To extract the interactions between the characters, two methods are proposed: the frequency method of their joint occurrence in different text units and the construction of word embeddings, using which the decision is made based on the cosine similarity. To implement this method, the novel by A. Dumas "The Three Musketeers" is chosen. The assessment of the quality of the implemented approaches is carried out in comparison with the answers of the respondents. The method of frequency co-occurrence in paragraphs for the main characters shows the highest intersection percentage of the top 3 agents with the answers of the respondents - 71%.