Emotional mining of prose works using a neural network (using the example of V.G. Korolenko's essay "Chudnaya")

Abstract

The article presents a methodology for conducting sentiment analysis of works of art using neural network technologies. The purpose of the research is to develop and test an algorithm for emotional mining of prose texts using the example of V.G. Korolenko's essay "Chudnaya". The material was the text of the work, divided into 26 logical blocks. The paper considers two models of analysis: binary ("positive–negative") and multicomponent, based on R. Plutchik's psychoevolutionary theory of emotions (eight basic emotions). The assessment of the tonality of words, expressions, and fragments was carried out using the Qwen neural network on eleven- and ten-point scales. The results are visualized in the form of tonality graphs, heat maps, and histograms of emotion distribution. It is established that the essay is characterized by emotional polyphony: all basic emotions are represented, but negative ones dominate — sadness and fear, which corresponds to the tragic plot. The tendency to increase the negative tone as the narrative develops is revealed. The method used allows for a quantitative assessment of the emotional dynamics of the text, the localization of the most intense passages, and their correlation with the events and character psychology described. The proposed approach demonstrates the possibilities of combining computational linguistics and literary studies to analyze literary texts, identify hidden patterns, and understand the author's intentions.