Inference Technique for Quantifying Political Leaning from Tweets

Abstract

The across the board utilization of online social networks (OSNs) to disperse data and trade assessments, by the overall population, news media and political performers alike, has empowered new roads of research in computational political science. In this Project, we think about the issue of evaluating and deriving the political inclining of Twitter clients. We define political inclining derivation as a curved enhancement issue that consolidates two thoughts: (a) clients are reliable in their activities of tweeting and retweeting about political issues, and (b) comparative clients have a tendency to be retweeted by comparable group of onlookers. We at that point apply our deduction method to 119 million race related tweets gathered in seven months amid the 2012 U.S. presidential race battle. On an arrangement of as often as possible retweeted sources, our procedure accomplishes 94% exactness and high rank connection as contrasted and physically made names. By concentrate the political inclining of 1,000 every now and again retweeted sources, 232,000 common clients who retweeted them, and the hashtags utilized by these sources, our quantitative investigation reveals insight into the political socioeconomics of the Twitter populace, and the transient progression of political polarization as occasions unfurl.