Using NLP Tools for Linking Materials Within the “Pushkin Digital” Resource

The development of the scientific and educational resource “Pushkin Digital” requires processing a substantial volume of documents to create an ontology of A. S. Pushkin's literary heritage. The works of A. S. Pushkin and related texts contain mentions of entities, such as historical figure names, geographical locations, dates, and references biographical sources. All these entities are the source of links in the ontology. The paper presents a description of the natural language processing techniques used in processing the materials featured on the Pushkin Digital resource in order to create links between them. The proposed system system utilizes state-of-the-art NLP techniques including BERT for robust named entity recognition, identifying and classifying key entities within the text, SBERT for enabling the system to discern relationships and connections between entities even when expressed with different wording, and LLMs for complex text analysis. Regular expressions are employed for identifying and processing structured text elements, such as dates and bibliographic references, ensuring data consistency and accuracy. This combination of techniques allows for the automated construction of a rich and interconnected ontology, facilitating indepth exploration of Pushkin's literary heritage and its broader cultural significance.