Sentiment Analysis to Assess Educational Methodologies in a Competency-Based Educational System

This paper presents the results of rule-based sentiment analysis to assess and adjust several interventional methodologies designed to develop students' mathematical and data-analytical competencies in mathematics, statistics, and data science e-learning courses. Regarding spatial mathematical skills, Tukey's simultaneous tests for the difference of means
$(\mathbf{p}=0.05)$
revealed that the Project-Based Learning methodology using 3D augmented reality and 3D printing obtained a significantly greater mean polarity than other methodologies. The same was true of data-analytical skills development using Research-Based Learning for Digital Storytelling and the Python programming language. Furthermore, feedback sentiment probabilities (positive, negative, and neutral) obtained from the rule-based sentiment analysis naturally defined an intensity-sentiment 3-dimensional model containing the feedback sample data. Persistent homology generated topological summaries of such feedback per semester and educational methodology in persistence diagrams. Amplitude vectors were calculated to quantify the closeness of the persistence diagrams.