Proposta de chatbot inteligente baseado na organização acadêmica do Instituto Federal de Pernambuco

Abstract

The Federal Institute of Pernambuco houses a variety of guiding documents. However, accessing the information contained within these documents is often neither simple nor quick due to their length and complexity. To address this challenge, this paper proposes the development of an intelligent chatbot designed to facilitate access to institutional information contained within the Academic Organization document of the Federal Institute of Pernambuco. The chatbot uses HTML, CSS, and ReactJS for the client interface, FastAPI for the server application, MySQL as the database, and the BERTimbau model for system intelligence. Additionally, the OrgAcadQA dataset was created, based on the Academic Organization document of the Institute, and used in conjunction with the SQuAD v1.1-PT-BR dataset for training and evaluating models in the Question Answering task. The BERTimbauLarge model achieved the most promising results, reaching an Exact Match of 0.78 and an F1 score of 0.88 on the OrgAcadQA dataset. These results highlight the effectiveness of BERTimbau models in building Question Answering systems within the context of the Academic Organization at the Federal Institute of Pernambuco.