Construction of a model for matching user's linguistic structures to a chat-bot language model

Abstract

The research object of this work is the linguistic structures of the user when constructing a request to a chatbot with generative artificial intelligence. The study solved the task of improving the communication mediation algorithms of chatbots through the comparison models of linguistic structures by users. Sometimes the user intentionally or due to lack of information forms an inaccurate request. Formally, this is described by logical operations "And" and "And or Not". As a result of the research, a model was built comparing linguistic structures at the input with the information model of the response at the output. The model was based on an approach with recursive creation of an answer. That has made it possible to determine the basic characteristics of the object of the request and form an answer on this basis. Using this approach improved the accuracy of the chatbot response. It also made it possible to consider the linguistic structure of the user through its formalization. The use of logic algebra made it possible to find typical errors of users during dialogs with generative artificial intelligence. A feature of the reported advancement is that the comparison of models of linguistic structures of query formation is carried out through a recurrent algorithm. As a result, it makes it possible to compare the query in such a way as to reduce the absolute error of the primary data by 0.02 % and simplify the process of mathematical calculations. At the same time, the received information becomes more accurate – the number of references increases from 2 to 6 sources. The proposal could be used in practical activities to improve the natural language recognition technologies of users in chatbots with generative artificial intelligence. On this basis, it is possible to devise various applications and services for training and practical activities