Development of a Question-answering System with Semantics-Oriented Natural Language Interface

The aim of this work was the development of a semantic question-answer system based on the natural-language interface. It was shown the need to create such automatic systems to reduce companies' costs for customer service. Under the tasks of this paper, the following items were selected: to consider the current state of the problem-response system, to consider methods for processing natural language, to choose one of the possible methods for creating similar semantic systems, and to test the possibility of introducing such a system into an existing project. In the first chapter, the definition of the question-answer system was given, examples were given of classifying such systems. It was shown that historically there was very much interest in these systems. After this, modern systems of this kind were shown and the modern problems faced by the developers of such systems were consecrated. The second chapter told about the methods of processing natural language. It showed two main approaches - statistical and linguistic. Linguistic approaches are also divided into two large classes of methods - semantic and morphological-syntactic. A class of SK-languages, proposed by V.A. Fomichev, which are based on the theory of K-representations. The next chapter was a continuation of the development of the latter theory. In this part, the possibilities of this theory were shown to be applied in practice. The last chapter describes the development of a semantically-oriented question-answer system in a natural language, providing processing of customer support issues for two different subject areas - banking and legal software products, and an attempt was made to implement this system in an already existing project. Despite the fact that the results were not good enough for the introduction of this product in real life, the opportunity was shown in the future to refine this system in a more efficient form and give recommendations for this.