Computing political preference among twitter followers

Abstract

Through anecdotal evidence and a variety of methods, claims are constantly being made about the bias of media outlets. As many of those outlets create online personas, we seek to measure the political preferences of their audience, rather than of the outlet itself. In this paper, we present a method for computing political preferences of an organization's Twitter followers using congressional liberal/conservative ADA scores as a seed. We apply this technique to characterize the political preferences of major news media Twitter audiences. We discuss how this technique can be extended or used to create personalized recommendations and insights for news seekers and social media analysts and marketers.

Stance and Sentiment in Tweets

“…There has been considerable interest in analyzing political tweets towards detecting sentiment, emotion, and purpose in electoral tweets , determining political alignment of tweeters [Golbeck and Hansen 2011;Conover et al 2011a], identifying contentious issues and political opinions [Maynard and Funk 2011], detecting the amount of polarization in the electorate [Conover et al 2011b], and even predicting the voting intentions or outcome of elections [Tumasjan et al 2010;Bermingham and Smeaton 2011;Lampos et al 2013].…”

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

“…However, there has been limited work on inferring the bias of content of short social media posts like tweets. Instead, researchers have inferred the bias of the users posting tweets by modeling how different polarity users use language (Purver and Karolina 2015;Makazhanov and Rafiei 2013;Fang et al 2015), or by leveraging the linking behavior of users (Golbeck and Hansen 2011;Conover et al 2011a, b), or by leveraging both textual and network features for political leaning classification (Pennacchiotti and Popescu 2011). 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.…”

“…Also, in Sect. 2, we have briefly described prior work which has developed techniques for measuring the bias of users (Purver and Karolina 2015;Makazhanov and Rafiei 2013;Fang et al 2015;Golbeck and Hansen 2011;Conover et al 2011a, b;Pennacchiotti and Popescu 2011;Bond and Messing 2015;Wong et al 2016) or content (Zafar et al 2016;Weber et al 2013) on social media as well as blogs and news stories (Adamic and Glance 2005;Yano et al 2010;Zhou et al 2011;Budak et al 2016;Munson et al 2013b) on the Web. In the future, when bias quantification schemes are developed for other search systems, for instance for videos (e.g., Youtube search) or music (e.g., Spotify), these methodologies can be plugged into our bias quantification framework and be used to analyze the bias of these other search systems.…”

Semantic Role Labeling of Emotions in Tweets

“…We use tweets from the 2012 US presidential elections as our dataset, since we expect political tweets to be particularly rich in emotions. Further, the dataset will be useful for applications such as determining political alignment of tweeters (Golbeck and Hansen, 2011;Conover et al, 2011b), identifying contentious issues (Maynard and Funk, 2011), detecting the amount of polarization in the electorate (Conover et al, 2011a), and so on.…”