Rocha, Fabio GomesRodriguez, Guillermo HoracioAndrade, Eli Emanuel F.Guimarães, AdolfoGonçalves, VitorFerraz Sabino, Rosimeri2021-12-072021-12-072021Rocha, Fabio; Rodriguez, Guillermo Horacio; Andrade, Eli Emanuel F.; Guimarães, Adolfo; Gonçalves, Vitor; Ferraz Sabino, Rosimeri (2021) Supervised Machine Learning for Automatic Assessment of Free-Text Answers. In Batyrshin I.; Gelbukh A.; Sidorov G. (Eds.) 20th Mexican International Conference on Artificial Intelligence, MICAI 2021. Mexico City: Springer Science and Business Media Deutschland GmbH. ISBN 978-303089819-9978-3-030-89819-90302-9743http://hdl.handle.net/10198/24469The learning assessment seeks to collect data that allows for identifying learning gaps for teacher decision-making. Hence, teachers need to plan and select various assessment instruments that enable the verification of learning evolution. Considering that a more significant number of evaluation instruments and modalities increase the teachers’ workload, the adoption of machine learning might support the assessing actions and amplify the potential of students’ observation and follow-up. This article aims to analyze machine learning algorithms for automatic classification of free-text answers, i.e., evaluating descriptive questions written in Portuguese. We utilized a dataset of 9981 free-text answers for 17 questions. After pre-processing the data, we used eight classification algorithms. In conclusion, we highlight that the Logistic Regression, ExtraTrees, Random Forest, and Multi-layer Perceptron algorithms obtained results above 0.9 of F-score for both multi-class and binary classification.engLearning assessmentSupervised machine learningMulti-class classificationFree-text answersTeacher decision makingconference object10.1007/978-3-030-89820-5_1