Гібридне виведення знань в інформаційних системах з використанням великих мовних моделей та графів знань

Abstract

This work is devoted to the study of hybrid knowledge inference in intelligent information systems using knowledge graphs and large language models. Modern information systems require effective methods for knowledge representation and reasoning to support intelligent data analysis and decision making. The main approaches to knowledge representation based on knowledge graphs and Semantic Web ontologies were analyzed. The possibilities of using large language models for processing natural language queries were also considered. A hybrid approach that combines symbolic reasoning based on knowledge graphs with neural methods based on large language models was studied. A comparison of symbolic, neural, and hybrid knowledge inference approaches was carried out. The results show that the hybrid approach can improve the flexibility and effectiveness of knowledge inference in intelligent information systems.