Naive Bayes Model with Improved Negation Handling and N-Gram Method for Sentiment Classification

Abstract

Sentiment classification is turning into one of the most fundamental research areas for prediction and classification. In Sentiment mining, we basically try to analyse the results and predict outcomes that are based on customer feedback or opinions. Some work has been done to increase the accuracy of the Naive Bayes classifier. In this project we have examined different methods of improvising the accuracy and space of a Naive Bayes classifier for sentiment classification. We have used a modified negation handling method using POS tagging to decrease the number of feature in the feature set and also discovered that combining these with n-gram method results in improvement in the accuracy. So, a more accurate sentiment classifier with less space complexity can be built from Naive Bayes Classifier.