METHODS OF CONCEPT EXTRACTION IN LITERARY WORKS

Abstract

The article explores both traditional and contemporary approaches to concept extraction in literary texts, with a particular focus on lexical analysis, semantic parsing, ontological modeling, and automated techniques based on large language models. Traditional manual methods are valued for their ability to capture nuanced literary meanings – such as symbolism, metaphor, and intertextual references – while taking into account cultural and stylistic context. However, their reliance on intensive human interpretation makes them difficult to apply to large text corpora and comparative studies. Modern automated approaches, especially those utilizing transformer architectures like BERT, introduce significant advantages in processing speed and scalability. Through mechanisms such as self-attention, these models effectively identify long-range contextual relationships and latent patterns within texts, enabling rapid detection and classification of key concepts across extensive datasets. Yet, the article emphasizes that these systems still face limitations when dealing with the figurative richness and semantic ambiguity inherent in literary discourse. The practical component of the research involves an experimental analysis of the concept of tolerance in articles from major English-language media outlets, including The New York Times, BBC News, and The Guardian. Automated extraction methods demonstrated strong potential for identifying general trends and conceptual usage patterns. Nonetheless, the findings underscore the need for expert interpretation to refine outputs, especially in cases involving subtle semantic shifts or interdisciplinary cultural references. In conclusion, the article proposes an integrative methodological framework that combines automated processing with human expertise. Automated tools are recommended for the preliminary structuring and classification of large volumes of textual data, while expert scholars are responsible for interpretative depth and validation. Future research directions include enhancing LLM adaptation for literary texts, building specialized training corpora, and incorporating ontological models to improve conceptual precision and reliability in literary studies.