TAKEI NETO, D. E.; http://lattes.cnpq.br/4469956777381477; TAKEI NETO, Diego Eizi.
Resumo:
Computational thinking (CT) is a reasoning process that consists of formulating a problem and solving it in steps that a computer is capable of solving. This process is so important that authors consider it as an enhancer of the competences operational aspects of human beings, which can be used in some strands, such as for example, in interdisciplinary development in basic education subjects, such as Mathematics and Physics, and in development from specific disciplines of Computer Science. In the context of the Mathematics discipline, we can relate an issue among nine computational thinking competencies. Identifying issues that explore these competencies can be extremely useful for students and teachers who have an interest in going deeper into this topic, as the stimulus to CT can increase the ability to solve problems. In this context, the design of intelligent models capable of predicting automatically CT skills in math issues would be a great facilitator in the process of stimulating the resolution of problems. In this work, we use a new database of questions to extract features from excerpts from the text assigned to questions by manual assessments from experts. From these questions, we developed
classifiers using Term Frequency-Inverse Document Frequency (TF-IDF) and Latent Dirichlet Allocation (LDA) as model features, recalculating the values of these characteristics through the use of the excerpts, with the objective to increase the importance of those that belong to the excerpts highlighted by the evaluators, and increasing the effectiveness of the classification of the questions in relation to the stimulated computational thinking competencies.