DEVELOPMENT OF A METHOD FOR DETERMINING THE EMOTIONAL STATE OF A PERSON USING A CHAT BOT

Abstract

Currently, dialogue systems (chatbots) are being intensively used in almost all areas of activity,
and it is worth noting that these mechanisms have begun to be used not only in the entertainment
sector, but also in more important areas - business or even medicine. In this regard,
there is a need to develop and improve the communication system of such robots in natural language.
This article proposes a method for determining a person’s emotional state using a chatbot.
The article defines a chatbot, discusses the types of chatbots, and defines the basic principles of
their work. Based on the identified differences in the operation of algorithms for text detection by
chatbots, the most suitable technology for solving the task is described - working with neural networks.
The shortcomings of existing solutions in comparison with the method being developed are
identified and described. It is important to note that the peculiarity of the method is the combination
of technologies of intelligent chatbots and augmented reality. The article describes the general
algorithm of the method for determining the emotional state of a person using a chatbot,
which includes eight main stages. The developed and identified criteria for assessing emotions are
described: valence, intensity and sthenicity. The architecture of the developed solution, which has
a client-server approach, is presented. The server part also includes a text sentiment detection module and a response generation module, while the client includes the chat and the appearance
of the “Avatar” of the developed chatbot. The main implementation steps are described, including
the initial interface, as well as the stages of developing an augmented reality application.
The article lists the tools used for working with augmented reality (Unity 3D and the AR Vuforia
framework). The created 3D models are presented. The results of processing skeletal animation
using Perception Neuron 3 equipment and subsequent processing in Axis Studio are shown. Based
on the results of the work, a conclusion was formulated describing the qualitative difference of the
developed system in relation to the existing analogues considered, and the results of testing the
operation of the system were also presented.