This paper describes the process of creating a knowledge graph for a collection of mathematical articles in the Russian language, gathered from the 'Izvestiya VUZov. Matematika' journal. The collection consists of approximately 1100 documents in LaTex format. The work involves constructing an ontology for the collection of mathematical articles, which will serve as the basis for the created knowledge graph. Various article objects are extracted from the collection, including universal decimal classification codes, authors, titles, used formulas, articles publication dates, authors affiliations and references to other works. Each object is recorded through a specific relationship in the knowledge graph. Thematic modeling is also performed on the collection using the latent Dirichlet allocation method, for which optimal hyperparameters are selected. The document themes are recorded in the knowledge graph through relationships. An interesting approach is used for extracting mathematical terms. In this work, mathematical entities are identified in the documents using the OntoMathPRO ontology. During the knowledge graph construction process, tools were developed that allow the creation of a knowledge graph on any collection that meets the patterns of the original collection. The resulting knowledge graph can serve as a foundation for various research purposes and the development of intelligent systems, that can be used by researchers, journals, as well as students.

Knowledge graph construction, Linked Data, Topic modeling, Mathematical Paper

Gizatullin Bulat Timurovich,  ,  orcid.org/0009-0000-6251-9260,  Kazan Federal University

Nevzoova Olga Avenirovna,  ,  orcid.org/0000-0001-8116-9446,  Kazan Federal University