М. В. Гончаров The purpose of the article is to present the research directions of the Russian National Public Library for Science and Technology (Russian NPLSaT) on implementing artificial intelligence (AI) elements into scientific and technical information systems. The focus is on two key directions: semantic transformation of user queries and automatic summarization and annotation of full-text materials. Approaches to normalization, enrichment, and structuring of queries using AI technologies to improve the completeness and accuracy of search in library catalogs and open archives are analyzed. Content and formal requirements for automatically generated abstracts and annotations are discussed, along with quality evaluation metrics (ROUGE, expert assessment). The paper presents experimental results comparing large language models (MBart, T-Bank) on datasets from the journal “Scientific and Technical Libraries” and environmental information collections of Russian NPLSaT. The conclusion emphasizes the necessity of synthesizing traditional library competencies with innovative AI technologies to ensure the competitiveness of scientific libraries in the new digital reality.
Научная статья 
Государственный рубрикатор научно-технической информации
scientific and technical library
semantic query processing
automatic summarization
automatic annotation
open archives
large language models
quality metrics
Russian National Public Library for Science and Technology
SRSTI
научно-техническая библиотека
семантическая обработка запросов
автоматическое реферирование
автоматическое аннотирование
открытые архивы
большие языковые модели
метрики качества
Государственная публичная научно-техническая библиотека России