Marchenko O. Semantic-Syntactic Models of Natural Language. Tensor approach

The thesis deals with developing semantic-syntactic model of natural language. The tensor approach to modeling semantic-syntactic relations between words in sentences was applied. Applying the apparatus of natural language syntactic structures control spaces allowed improving the tensor semantic-syntactic model. Using recursion and superposition in the tensor model allows describing syntactic patterns of any length and complexity. After factorization of constructed tensors the system generates vectors of semantic-syntactic valence of words. They describe the commutative behavior of the words in natural language. A method for calculating vectors of semantic-syntactic valence for ontology concepts that are implicit description of their semantic relations has been developed. Vectors of semantic-syntactic valences of words have been used as context vectors for formal concept analysis algorithm that allowed the automatic generation of high-quality taxonomies. A basic algorithm for automatic construction of ontological knowledge base has been developed on the basis of the tensor semantic-syntactic model of natural language.