The paper aims to address the studies of operating scientific databases using AI. The main goal is to organize queries in a way that is easy to understand, avoiding the use of structured queries or database search services. An overview of the main trends in this area, with a general assessment of their potential, has been carried out. Special attention is given to the use of large language models for querying non-relational databases. The results of our research on large language models for querying a graph database containing an ontology of the 'thermophysical properties of substances' domain are presented.

natural language query, large language model, embedding, no relational databases, graph database, domain ontology