Baseline evaluation: an empirical study of the performance of machine learning algorithms in short snippet sentiment analysis

摘要: At present, the sentiment analysis task is a relatively recent research field which has led to many inconsistent findings in literature. The debate two-fold: what best performing baseline classifier and most useful feature weighting method e.g., term presence (TP), frequency (TF), TF-IDF etc., can be used improve classifier's performance. Naive Bayes, with its variations Support Vector Machine are commonly task. However, their reported performance varies among researchers divergence as that In order shed some light on this controversy, we have conducted series of widely comparative experiments (including twelve various domains) evaluate machine learning classifiers (Naive Bayes variations, J48 - an implementation decision tree-based -) experimental results indicate Binarized Multinomial (BMNB) exhibits short snippet Furthermore, classification performance, using selection methods, namely information gain (IG), been significantly improved.