Polarity Classification of Twitter Data using Sentiment Analysis

Abstract

Crowd source information is of vital importance these days, since we relay much on information available from internet. Thus, Sentiment analysis or opinion mining becomes one of the major tasks of NLP (Natural Language Processing) and has gained much attention in recent years. Sentiment analysis is the process of computationally identifying and categorizing opinions expressed in a piece of text, specifically to determine whether the user’s attitude towards a specific area or product in case of ecommerce, etc. is positive, negative, or neutral. Sentiment analysis application are broad and powerful. It can be helpful in many ways like it helps marketers to evaluate the success of an ad campaign, in new product launch, to determine which versions of a product or service are popular and it also identifies which demographics like or dislike product features. This paper evaluates two classifiers, one is linear and other is probabilistic for sentiment polarity categorization. Data used in this study are the tweets collected from twitter.com. We further represent a comparative study of three different algorithms, Naïve Bayes, SVM (Support Vector Machines), and Logistic regression and how they vary on the same data set.