Models and methods for automatic comparative semantic analysis of scientific texts

The article considers a general approach to the implementation of algorithms for comparative analysis of scientific texts from the position of their general content and from the position of dynamic sequence of text fragment semantics. The proposed algorithms for comparative analysis of scientific texts are based on the clustering of semantic graphs, which are constructed as a result of combining information extracted from the publication with the scientific domain semantics. Approaches to extracting a dynamic subsequence of semantically similar fragments are discussed, and ways of calculating correlation coefficients are given. The paper presents experimental results of static and dynamic comparative analysis of mathematical publications, which are graphically illustrated.