Andrushchenko, D.Klimenko, V.Mazurets, O.Мазурець, Олександр Вікторович2025-07-012025-07-012025Andrushchenko D., Klimenko V., Mazurets O. Vector Databases Search for Adaptive Filtering of Scientific Articles. Scientific trends in the development of modern technologies. Proceedings XXII International Scientific and Practical Conference. May 28-30, 2025. Krakow, Poland. Pp. 189-195https://elar.khmnu.edu.ua/handle/123456789/18991The paper explores a modern approach to adaptive filtering scientific articles by integrating neural network-based text embeddings with vector database architectures. In response to academic publications’ exponential growth and semantic complexity, the proposed system leverages pre-trained language models to convert scientific texts into high-dimensional vectors and stores them in a vector database optimised for similarity search. This enables personalised and context-aware recommendations based on user profiles formed through explicit preferences and implicit behaviour. The modular architecture includes subsystems for data ingestion, text preprocessing, vector generation, model training, and adaptive retrieval. The core advantage lies in the semantic matching of content beyond traditional keyword-based methods, enabling researchers to access relevant publications even across interdisciplinary domains efficiently. Challenges related to computational cost, domain adaptation, and explainability are acknowledged as key areas for further research. The study underscores the practical importance of improving scientific recommender systems for enhancing knowledge accessibility and research productivity in the digital era.enСтаття