Classificação de Textos Escolares com Aprendizado de Máquina

Abstract

RESUMONowadays, during the teaching-learning process, it is prevalentfor students to seek information about the subjects to complementstudies or execute school work. For this, it is possible to use searchengines on the Internet, but the texts do not always have the complexitycompatible with the student’s level of education. A solutionfor this purpose is the automatic filtering of texts, where computationalmodels are created for textual classification according totheir contexts. Many studies have presented solutions based onMachine Learning techniques, but there is no comparison nor ananalysis considering these methods. This work explores this scenariothroughout implementing machine learning algorithms. In thesolution, Natural Language Processing techniques are applied inorder to select characteristics of the textual contents and introducethem into the classification pipeline. Experiments were conductedand the results obtained suggest that, machine learning algorithmshave a good predictive performance values, with class-balancedaccuracy of around 0.89 in the test data.