Exploiting Ordinality in Predicting Star Reviews

Automatically evaluating the sentiment of reviews is becoming increasingly important due to internet growth and increasing customer and business use. We hope to address the question of what is the best model for classifying a review’s text to its labels. We propose using a classifier that combines metric labelling and ordinal regression. Our results showed that metric labeling was not improved by combining it with ordinal regression. Moreover, our results indicate that a one-vs-all classification approach may be best way to classify reviews.