Authorship detection using graph models based on anthroponyms (on the material of Russian-language fanfiction)

Fan fiction is a potentially rich field for the study of both the writer's linguistic creativity through the anthroponymic space of the work, and all kinds of textual features identifying the author's style. While solving these tasks through a linguistic analysis, the manual work of an expert remains the main limiting factor, which creates the need to minimize the number of actions through automation. One of the ways to improve the process of a linguistic analysis, in addition to using machine learning algorithms, is to represent the data in the form of graphs that are can display complex relationships between objects. On account of this, the use of graph structures has increasingly been gaining popularity in the tasks of formal text attribution. The purpose of this work is to apply graph models, including those using the anthroponymic space of the texts, for attribution of fanfics. In addition, the texts are also examined for occasional anthroponyms, their characteristics, and onomastic techniques. The modern methods of Natural Language Processing (graph neural networks, named entity recognition systems) serve as the main tools for achieving the set goals.