Opinion mining of customers reviews using new Jaccard dissimilarity kernel function

Abstract

Opinion mining (aka sentiment mining), a subdivision of text classification has become traction among researchers in recent decades, due to the popularity of its practical application in real-time scenarios like product reviews, politics, movies, etc. Various machine learning algorithms are used to identify the document or sentence opinions which are available in social space. SVM is one of the most popular supervised machine learning algorithms and uses kernel function to classify data when the data points are nonlinearly separable. In this paper, we have proposed a new Kernel function called Jaccard dissimilarity Kernel functions where the distance between the two binary vectors is classified based on principle of Jaccard coefficient. In our study, we used this Jaccard Kernel function to classify the opinions of the recent Bollywood movie reviews in to positive and negative.