APPLICATION OF LARGE LANGUAGE MODELS FOR BUILDING A “FOREST OF TERM HIERARCHIES”

Abstract

One of the ways to organize and systematize knowledge is to form terminology ontologies that allow you to structure information in specific subject areas, such as cybersecurity. With the revolutionary emergence of large language models (LLMs), new opportunities are emerging to automate the process of building a "forest of term hierarchies" (FTH). Building a FTH is essential for several key aspects of cybersecurity and knowledge management, such as unifying terminology, improving communication, optimizing information retrieval, systematizing knowledge, adapting to new challenges, and supporting research and innovation. The article discusses the role of LLM in building FTH in the context of modern challenges of the information environment. Thanks to revolutionary advances in artificial intelligence, LLMs automate and optimize the processes of processing, analyzing, and structuring large amounts of text data. The article describes the key stages of FTH implementation using LLM, including text data processing, determining the discriminant power of terms, establishing links between them, and visualizing the results. A methodology for determining the associative relationships between predefined terms for building FTH is proposed. Examples of the practical implementation of the proposed methodology based on the use of the information-analytical system "Cyber Aggregator" are given. An example of forming a prompt for building a FTH for the generative artificial intelligence system DeepSeek.com is demonstrated. The technology of FTH visualization is proposed by using the program for analyzing and visualizing graphs CSV2Graph. The use of the proposed technologies makes it possible to increase the efficiency and accuracy of building terminological ontologies, which is important for adapting to the rapidly growing information flows in the modern world.