UDC 007:681.512.2
THEORY OF HIERARCHICAL NUMBERS IN CALCULATION PROBLEMS SEMANTIC SIMILARITY OF NATURAL LANGUAGE CONSTRUCTIONS
I. Yu. Kashirin, Dr. in technical sciences, full professor, RSREU, Ryazan, Russia;
orcid.org/0000-0003-1694-7410 , e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
The algebra of hierarchical numbers, operations and relations of an algebraic system are considered. A graphical representation of hierarchical numbers and the operations with them is provided; distinctive properties of the operations are shown. Methods for normalizing hierarchical numbers for their subsequent use in processing natural language constructs are listed and explained. To use the theory of hierarchical numbers, knowledge models ontologies are developed in terms of generic taxonomies, which also have hierarchical structure. General and applied ontologies having significant differences in their design and application for understanding natural language sentences are distinguished. As a cross-cutting example, we took the subject area of English-language political articles of international electronic media, in particular: RT, CNN, TASS, NYTimes. The technology for calculating the semantic similarity of natural language constructions is considered, for which well-known bert-base-cased neural network models of the latest versions are used, as well as the author’s IYu-bert-cased model. A new method for computing semantic similarity using hierarchical number theory is presented. The experimental part of the material is based on the use of software tools of Python v.3 language (Anaconda 3): Spacy library v.3.2.1, CorpusMining v.2.1 retriever, mIYu-bert v.1.0 software package. The last two tools were implemented by the author of the material. The completed series of experiments allows us to qualify the methodology for using hierarchical numbers in calculating semantic similarity as the basis of the technology that is not inferior in efficiency to currently available international analogues. The aim of the work is to present the effective use of hierarchical number algebra to obtain and use new neural network technology used to solve the problems of automatic calculation of semantic similarity in natural language constructions.
Key words: : hierarchical number theory, neural Bert models, natural language analysis, ontological taxonomies, semantic similarity