Desenvolvimento de um chatbot educacional para apoio aos estudantes do curso de Engenharia de Software da UFC-campus Quixadá

Abstract

This work presents the development of an educational chatbot aimed at providing academic support to students of the Software Engineering program at the Federal University of Ceará– Quixadá Campus. The project emerged from the need to automate responses to recurring questions faced by students, streamlining communication with the course coordination and promoting greater autonomy in accessing institutional information. The chatbot was designed to answer inquiries related to academic procedures such as mandatory internships, course withdrawal, credit hours, course equivalency, and elective subjects. To achieve this, modern Natural Language Processing (NLP) techniques were employed, notably the use of semantic embeddings, symbolic classification, and a fallback mechanism for unrecognized queries. The hybrid architecture developed enables the interpretation of various linguistic formulations used by students, ensuring greater accuracy in intent recognition. The solution was implemented using the Hugging Face Transformers library and deployed via the Telegram platform. Experimental evaluation with real users showed a high level of accuracy, satisfaction, and usability, confirming the technical feasibility and practical relevance of the tool as complementary support for academic management.