Twitter and polls : analyzing and estimating political orientation of Twitter users in India General #Elections2014

Abstract

This year (2014) in the month of May, the tenure of the 15th Lok Sabha was to end and the elections to the 543 parliamentary seats were to be held. With 813 million registered voters, out of which a 100 million were rst time voters, India is the world's largest democracy. A whooping $5 billion were spent on these elections, which made us stand second only to the US Presidential elections ($7 billion) in terms of money spent. The di erent phases of elections were held on 9 days spanning over the months of April and May, making it the most elaborate exercise to choose the Prime Minister of India. Swelling number of Internet users and Online Social Media (OSM) users turned these unconventional media platforms into key medium in these elections; that could e ect 3-4% of urban population votes as per a report of IAMAI (Internet & Mobile Association of India). Political parties making use of Google+ Hangout to interact with people and party workers, posting campaigning photos on Instagram and videos on YouTube, debating on Twitter and Facebook were strong indicators of the impact of the OSM on the General Elections 2014. With hardly any political leader or party not having his account on the micro blogging site Twitter and the surge in the political conversations on Twitter, inspired us to take the opportunity to study and analyze this huge ocean of elections data. Our count of tweets related to elections from September 2013 to May 2014, collected with the help of Twitter's Streaming API was close to 18.21 million. We analyzed the complete dataset to nd interesting patterns in it and also to verify if the trivial things were also evident in the data collected. We found that the activity on Twitter peaked during important events related to elections. It was evident from our data that the political behavior of the politicians a ected their followers count and thus popularity on Twitter. Yet another aim of our work was to nd an e cient way to classify the political orientation of the users on Twitter. To accomplish this task, we used four di erent techniques: two were based on the content of the tweets made by the user, one on the user based features and another one based on community detection algorithm on the retweet and user mention networks. We found that the community detection algorithm worked best with an e ciency of more than 80%. It was also seen that the content based methods did not fare well in the classi cation results. With an aim to monitor the daily incoming data, we built a portal to show the analysis of the tweets of the last 24 hours.1 This portal analyzed the tweets to nd the most trending topics, hashtags, the kind of sentiments received by the parties, location of the tweets and also monitored the popularity of various political leaders and their parties' accounts on Twitter. To the best of our knowledge, this is the rst academic pursuit to analyze the elections data and classify the users in the India General Elections 2014.