Shuklin D. Semantic neural network models and their application in the artificial intelligence systems

Thesis for a candidate's degree of technical sciences on speciality 05.13.23 - the Artifi-cial Intelligence Systems and Means. - the Kharkov National University of Radio Electronics, Kharkov, 2003. The dissertation is devoted to the solution of a symbolical sequences identification problem, definitions of their morphological and syntactic attributes, processes of synonymy and homonymy phenomena in the limited natural language texts. The developed dissertation presents: the semantic neural network model which allows to represent formal model of the text as algebra with operations carried by neurons on the natural language elements; the syn-chronized linear tree model as a semantic neural network connections structure being the fi-nite automata and providing: identification of the lexemes, morphological and syntactic at-tributes corresponding to the text elements and processes of the morphological and syntactic synonymy and homonymy; the line of time model carrying out the function of short-term memory;the model of direct clause conclusion mechanism as a structure of neural connec-tions, carrying out the semantic analysis of the text elements. A specially modified semantic neural network processing synonymy and homonymy of financial parameters in electronic documents has been installed in the executive office of Kharkiv regional department of So-cial Assurance Fund for temporary disability, and it has received a trade mark "Yes, Іt works Offіce Extensіons" from the PC Magazіne Russіan Edіtіon Test Laboratory and has been placed in the Mіcrosoft Offіce Extensіons library.