Quantifying Political Leaning from Tweets, Retweets, and Retweeters

Abstract

The widespread use of online social networks (OSNs) to disseminate information and exchange opinions, by the general public, news media and political actors alike, has enabled new avenues of research in computational political science. In this paper, we study the problem of quantifying and inferring the political leaning of Twitter users. We formulate political leaning inference as a convex optimization problem that incorporates two ideas: (a) users are consistent in their actions of tweeting and retweeting about political issues, and (b) similar users tend to be retweeted by similar audience. Then for evaluation and a numerical study, we apply our inference technique to 119 million election-related tweets collected in seven months during the 2012 U.S. presidential election campaign. Our technique achieves 94% accuracy and high rank correlation as compared with manually created labels. By studying the political leaning of 1,000 frequently retweeted sources, 230,000 ordinary users who retweeted them, and the hashtags used by these sources, our numerical study sheds light on the political demographics of the Twitter population, and the temporal dynamics of political polarization as events unfold.

Search bias quantification: investigating political bias in social media and web search

“…Zafar et al (2016), have quantified the impartiality of social media posts by measuring how easy it is to guess the political leaning of its author. Bond and Messing (2015) inferred the political leanings of Facebook users by observing the endorsements of Facebook Pages of known politicians, while Wong et al (2016) measure the endorsements in terms of retweeting behavior of users to infer their political leanings. Cohen and Ruths (2013) used supervised methods to classify users into different groups of political activities and showed that it is hard to infer the political leaning of "normal" users.…”

Aligning Entity Names with Online Aliases on Twitter

“…Many build learning models that are applicable across a variety of different user attributes (Chen et al, 2015a;Volkova et al, 2015;Beretta et al, 2015). Among the attributes, political preferences is a frequent area of research, again relying on features from user tweets, and making use of graph-based algorithms over their friends' attributes (Golbeck and Hansen, 2011;Conover et al, 2011;Wong et al, 2013;Cohen and Ruths, 2013;. Pla et al (2014) even uses sentiment analysis.…”

Proceedings of the Second Workshop on NLP and Computational Social Science

“…Characterizing social media users based on their political affiliation is an ongoing challenge in Natural Language Processing and Computational Social Science (Conover et al, 2011;Cohen and Ruths, 2013;Sylwester and Purver, 2015;Wong et al, 2016). In addition, linguistic reflections of political identity are of interest to sociolinguists (Hall-Lew et al, 2010;Labov, 2011).…”

“…Sylwester and Purver, who were interested in characterizing psychological differences between Democrats and Republicans, focused on word frequency, friend-follower ratio and Linguistic Inquiry and Word Count (Pennebaker et al, 2001)-although they also excluded punctuation from their data. Another study by Wong et al used no linguistic features at all, relying instead on social network relations with users whose political affiliation was known (Wong et al, 2016). Much of the sociolinguistic work, on the other hand, has focused on sub-lexical features that encode political identity.…”