Accuracy' Measures of Sentiment Analysis Algorithms for Spanish Corpus generated in Peer Assessment

The purpose of this study is to test a model that classifies some sentiment as positive or negative from some feedback in Spanish that are generated through peer assessment in Higher Education. The Supervised Machine Learning method is implemented. Several experiments are performed with a manually tagged data set to test different combinations of N-grams with Term Frequency-Inverse Document Frequency (TF-IDF), and classification algorithms: Multinomial Naive Bayes, Support Vector Machine, Logistic Regression, and also Random Forest, in order to obtain the right combination that gives the best performance. The simulation results displayed that the Support Vector Machine classifier with the combination of 1-grams + 2-grams + TF-IDF is the best model in Precision, Recall and F-Measure.