AVALIAÇÃO DE REPRESENTAÇÕES VETORIAIS NA CLASSIFICAÇÃO DE QUESTÕES DE VESTIBULAR POR COMPETÊNCIAS: UM ESTUDO COMPARATIVO

Abstract

 Identifying the skills needed to solve university entrance exam questions can be challenging for students, both due to the complexity of this process and the high volume of available questions. Therefore, automating the classification of questions by competency presents itself as a means of assisting students. As far as we have observed, AI research focused on creating machine learning models to classify questions by competence is scarce, suggesting a disagreement about the types of embeddings to be used in this context. Therefore, this work proposed classifying questions by competency using different types of embeddings. The experimental results indicate that the use of simple embeddings can lead to classification models with performance comparable to or superior to models built from more sophisticated embeddings.

Keywords: Questions Classification; Competencies; Embeddings; Artificial Intelligence.