DD-CSS: Social Media Data Collector & Analyzer
DD-CSS is an open source project and a web application that facilitates data collection from social media and provides tools for data analysis. Although several nice repositories are publicly available, to the best of our knowledge there is no web application for collecting & analyzing social media data that does not require any programming skills.
We have many tools for social network analysis such as Gephi, Pajek, UCINET, etc. but what about the data collection? NodeXL, to our knowledge is the only one that can directly pull data from social media sites but it only works in MS Windows and Excel. Also there are many other kinds of analysis on social media than what is offered by these systems.
Our goal is to build a social, crowdsourced fact checking website. There are several popular fact checking websites in the US such as snopes.com, factcheck.org, or politifact.com but the posts in these websites belong only to a few editors. The main challenge in this project is community moderation as in the case of Stack Exchange or Wikipedia.
Polarity of Turkish News Media
The Turkish news media is believed to have a high degree of polarized pluralism and political parallelism. In this project, the extent of this belief is computationally investigated at a large scale by collecting more than ten million Twitter user IDs. We compiled two datasets, the first being the news audience dataset which is composed of user IDs who follow at least one popular Turkish news media among thirty-seven media accounts selected. The second dataset relates to the political audience which consists of IDs following Twitter accounts of any of the four political parties in the current Turkish parliament. We first measure the pairwise similarities of news media based on their common followers and then detect the media groups at different resolutions and finally represent their relative positioning in two dimensional space. It is observed that media positioning and clusters are well aligned with the known ideologies of the media groups. We then measure the polarity of the news audiences’ political leanings and also investigate the news media preference of the party followers. Through such analysis we show that the media preferences of parties are quite different from one another and party descriptiveness of the media is almost completely reverse ordered for the parties in the opposite camps. Finally, to highlight these findings interactive visualizations are also created to make findings easier to interpret by a broader audience.
Publications: PolNet 2014