Desenvolvimento de um chatbot para atendimento público

Abstract

The higher education admissions process attracts a significant volume of candidates annually, generating a high demand for information. The overloading of traditional support channels results in long wait times and the repetition of answers to frequent questions, compromising the quality of the candidate experience. This paper presents the design and development of a chatbot prototype to automate and qualify public support, focusing on queries about the selective processes at UTFPR, with an initial emphasis on the vestibular (entrance exam). The adopted methodology involved the implementation of a microservices architecture. The solution employs the Rasa framework for natural language processing and dialogue management, while a backend developed in Python using the FastAPI framework, integrated with the Qdrant vector database, performs semantic searches to retrieve the most pertinent answers. The results demonstrate a functional system composed of a chatbot capable of interpreting user intent and a management interface that allows non-technical administrators to keep the knowledge base updated.