Sentiment Analysis of Twitter Textual Data Using Map Reduce on Demonetisation of Money in India

Abstract

The fast advancement and also utilization of the second stage of development of the Internet and Online networking has led the users generating huge volumes of data. A lot of sites on the web offer users to express their opinions on various products, people and events. This leads to an opportunity for mining sentiments from large unstructured data. In this project we implemented a dictionary based algorithm (which uses a predefined dictionary instead of a classifier to determine whether a word is negative, positive or neutral) on map reduce framework that is capable of processing large amount of data. A large amount of tweets are fetched using “Twitter Developer API” to HDFS. These tweets are then pre-processed to extract relevant information and then are further analysed to determine the sentiment of tweet. The output is the time series visualisation of average sentiment about the given subject. Using this system on the subject demonetisation of money in India lead to the observation that most people have neutral sentiment towards the issue followed by positive and then negative.