Visualization of the Public’s Opinion on Politically Influential Tweets

We present a Twitter-based website The Political Grid Project that focuses on harnessing the potential of Twitter as a political medium. On this site, the general public can create new datasets additionally to existing Twitter data, and investigate the data through diverse visualizations. This site lists tweets posted by political entities, and a user votes on each tweet based on the two standards—how much she agrees with the idea expressed in the tweet, and how important it is to her. More importantly, our site provides users with various interactive visualizations that represent both individual and collective voting data accompanied by data accessed from Twitter APIs. While exploring the visualizations, users can closely examine various aspects of politics, including their political preferences, the ties among the voted entities based on the analysis of the collective data, and temporal or geographical patterns. Ultimately, we expect our users to reflect on the political issues in tweets that are otherwise ephemeral and to better understand candidates in upcoming elections. Furthermore, our site allows users to investigate the public’s aggregated opinion and their political stance among the public. Background and Problems Twitter, the most popular micro-blogging service, has become a crucial medium for the contemporary politics. Politicians and political parties take advantage of Twitter to spread their voices. So do journalists and their news channels to response to these voices and drive the public’s opinion. In the vast Twitter-sphere, many third party applications exist that focus on these voices from politics and journalism. (e.g., washingtonpost.com/ mention-machine, wefollow.com/twitter/politics, 2012twit.com). These sites show influential Twitter users based on the number of followers and tweets that are freely acquired through Twitter APIs. Additionally, the APIs enable users to retrieve data about the spread of a single tweet including Copyright © 2012, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. the number of re-tweets and replies. However, these general descriptions do not adequately describe the complexity of the publics’ responses to the political voices and their tweets. For example, a follower of a political Twitter user may re-tweet or reply when she likes it as well as when she does not like it. We believe that what people really think about the tweet is more important. To our knowledge, there are no Twitter applications that enable end users to create new types of data, which can explicitly express the users’ opinion, besides consuming and adding to the Twitter datasets. Twitter data is a unique and interesting resource for research in social sciences (social media studies, journalism, and public policy) and engineering (data mining and information visualization). Researchers in these fields exploit the data to discover insights about human behavior and social networks that are politically meaningful; investigating the effects of Twitter on an election (Tumasja 2010), discovering the network segregation according to political orientations (Conover 2011), and computing the political preference of the citizens (Golbeck 2011) to list a few. Through complex qualitative and statistical methods, these projects demonstrate that Twitter can be a barometer to predict diverse phenomena in the creation of public opinions and the effects on related political events. However, these research projects were conducted with data previously obtained for academic purposes, and did not leverage the real-time participation of the public. The Political Grid Project The approaches of third-party applications and research projects are summarized as either re-presentation of the existing Twitter data or only pertaining to academic concerns. Three factors differentiate our project from the previously mentioned applications and projects. First, our site presents politically influential tweets, not the real time stream of Twitter user’s followings. Second, we do not AAAI Technical Report WS-12-03 Social Media Visualization