COMPUTER MODEL OF RESPONSE GENERATION IN THE SEARCH SYSTEM BASED ON AN UNSTRUCTED KNOWLEDGE BASE

Abstract

The aim of the work is to develop a request-response system model capable of creating specific textual responses to a user’s request, using a scientific text generating in a natural language in its algorithm. The system determines the semantic links in the documents, while creating a new text containing the answer to the user's question. The article discusses a model of a system based on the developed approach to the formation of a semantic model of a document, which allows you to get quantitative indicators of the semantic properties of a document in a natural language and semantic links between components of a scientific text. Developed as an application software, the semantic search system should use a model capable of working with a sufficiently formalized type of knowledge, namely - scientific text and allow you to automatically form a software semantic model of a single document and the body of knowledge as a whole. Based on the received structure, the application should generate a text response to the received user request. This leads to an important scientific property of the created model - the application should be able to use unstructured corpus of texts, which is an unstructured knowledge base, for which it is necessary to create and explore a semantic model of scientific text in natural language, and develop an algorithm for its formation from the semantic network. This approach solves most word processing issues for further automatic generation. In addition, a subsystem for the automatic classification of scientific texts by the degree of their connectivity was developed, which uses quantitative characteristics of the created model of a scientific text in its work. The article describes the developed criteria for evaluating the created systems and algorithms. The system thus obtained, in addition to organizing a convenient search environment, forms a universal model for automatic text processing at a semantic level for groups of Slavic-language texts of a formal style, a set of tools that allow you to flexibly create and process thematic full-text document bodies without preliminary semantic markup and get a program text model formalized stylistic orientation with quantitative characteristics of semantic properties text and, on the basis of which it is possible to solve other problems of automatic word processing.