Мікросервіс заповнення моделі знань термінів‑визначень із документів PDF формату

Abstract

The article describes a microservice for automated generation of ‘term-definition’ pairs from documents, in particular PDF files with complex internal structure. The processing pipeline includes file type identification, text extraction using the pdfplumber library, normalisation and artefact removal, segmentation, identification of defining structures based on templates and heuristics, and generation of results in JSON format for integration into educational systems. The development supports Ukrainian and English languages, has the prospect of expansion to other formats and integration with NLP models to improve semantic relevance. The advantages of the solution are transparent architecture, efficiency, stability when working with heterogeneous PDFs, and no need for additional model training.