N-Grams_effect_on_analysis_of_online_reviews_sentiments

Evaluating reviews for positive and negative subjective expression can be important in applications such as selecting a good product on online shop, selecting a nice movie or music on website, etc., there are three main problems: a large amount of reviews, create an optimal model, select appropriate feature set. In this paper, is employed three classifiers in the experiments: Passive-aggressive, Winnow, and Perceptron. And effect of n-grams as feature is evaluated on classifiers. In this paper, is compared perceptron classifier to winnow and passive- aggressive classifiers using n-grams features set. The performance for the passive- aggressive and the perceptron classifiers was excellent and close, but the performance for the perceptron classifiers using unigrams as feature was significant.