Bi Modal Sentiment Analysis of Gurmukhi Social Media Using Text Image Microblogs

Abstract

Everyone is active on social media in expressing their views and communicating with others due to the introduction of social websites and mobile applications i.e. Facebook, Twitter, Instagram, WhatsApp etc. whether there is election, current incidents, social conflicts, religious clashes, family visits etc. people express their views on social media. Out of these, twitter became more effective where one can express their feelings in a few words and post it as a text or along with picture also. It founds more interesting platform where people post their views and others comments on it which makes them related and interactive to each other. Their posts or comments not only express themselves but also expresses their moods, Emotions and present states of mind. Punjabi community especially eastern Punjab is also activated and communicates with each other in Gurmukhi script. Emotion detection in Gurmukhi has not been attempted that much yet and in this work, it has been carried out by collecting Gurmukhi data from Twitter Using Twitter API. Collected content has been filtered and six categories of data have been generated named as Sensitive, Happy, love, Religious, Sad and Angry moods of people. There were 4237 documents left after filtering. The proposed work was carried out in three phases: 1) Emotion analysis from text 2) Emotion analysis from Image 3) Bi-modal for Emotion analysis from image and text. In text based Emotion Classification, for Feature extraction, TF-IDF and N-gram features are used which are reduced further Using MI (mutual information) and Particle swarm optimization (PSO). As larger number of FeatureSet is reduced Classification speed of classifiers and the accuracy of Classification improves. FeatureSets having high AUC values are chosen and projected for Classification by three classifiers named as Decision tree, Naïve-Bayes and k nearest neighbor. It has been analyzed that by applying TF-IDF word level and N-gram and Feature reduction through MI-PSO, performance of Emotion Classification is ab