Порівняльний аналіз великих мовних моделей для чат-асистента у сфері ментального здоров'я

Abstract

This work presents a comparative analysis of modern large language models for the development of a chatbot assistant in the field of mental health. The study examines the ability of LLMs to interpret user queries expressed in natural language and generate responses that assist users in selecting appropriate mental health specialists. Several leading models are compared according to criteria relevant for dialogue systems, including general intelligence, instruction following ability, response style and naturalness, and API cost efficiency. Based on this analysis, a suitable model is selected for implementing a prototype chatbot system designed to assist users in navigating mental health services and identifying appropriate specialists.