Automating the Formation of the Conceptual Structure of the Knowledge Base Using Deep Learning

Introduction. The ability to automate processes is a key aspect of modern information technology. The construction and use of the conceptual structure of the knowledge base is becoming an urgent need in the modern world, where the amount of information is growing exponentially. The ability to automate processes, including the construction of ontologies, which requires the extraction of knowledge from full-text sources and their automatic structuring, is important. Knowledge bases are used to manage complex dynamic systems by ensuring the storage, organization, and access to a large amount of information that allows for effective analysis and prediction of the behavior of such systems. The purpose of the paper. The purpose of the paper is to demonstrate the effectiveness of using deep learning methods to automate the formation of the conceptual structure of the knowledge base. The study also aims to show how the integration of knowledge bases with deep learning methods can improve the quality of forecasts and increase the efficiency of rehabilitation trajectory management. Results. The algorithm successfully extracted and processed symptom information from the medical cases, effectively handling duplicates and synonyms. The utilization of cosine similarity enabled the identification of synonymous symptoms within the established knowledge base, facilitating the seamless integration of new information while preventing redundancy. The system demonstrated its capability to discern which symptoms should be incorporated into the knowledge base and which should be omitted based on their similarity to existing entries. The outcomes underscore the potential of this automated approach to enhance the knowledge base and contribute to the refinement of predictive models within the healthcare domain. Conclusions. The study demonstrated the effectiveness of deep learning in automating the formation of the conceptual structure of a medical knowledge base. The approach enhances the filling and comprehensiveness of the knowledge base, which is crucial for building predictive models for patient trajectories and improving healthcare decision support. Keywords: Knowledge-Oriented Management Systems, knowledge base, Support Vector Machine, Word2Vec, Skip-Gram, BioBERT.