Generación automática de dntologías de dominio específico a partir de textos académicos con técnicas de Deep Learning

Abstract

Abstract: The objective of this Degree Project (TT) is to develop a framework for the automatic generation of specific ontologies from academic text corpora, using deep learning techniques as a central component for knowledge extraction. The development of the TT focuses mainly on the analysis of abstracts, emphasizing abstracts due to their high quality of information, conciseness, and availability, thus establishing a flow that integrates several phases such as: corpus acquisition and curation, textual preprocessing, concept and relationship extraction, and the construction of ontological classes and properties using prompt engineering strategies. This research presents an alternative for ontological construction that stands out for its effectiveness in synthesizing a reproducible process to transform scientific information into reusable semantic representations. In this way, the organization and retrieval of knowledge in academic settings is facilitated, establishing the foundations for subsequent applications and studies with different purposes.