Loading…
This event has ended. View the official site or create your own event → Check it out
This event has ended. Create your own
View analytic
Saturday, September 27 • 10:41 - 11:00
"Iranian Political Landscape Seen from Iranian Presidential Election Tweets"

Sign up or log in to save this to your schedule and see who's attending!

Background: Social media plays an increasingly critical role in our political life – political parties and politicians use it to push their political agendas and advance their causes, and common people use it to share their political views and play an active role in elections and alike. This is witnessed by the recent and historical 2013 Iranian presidential election, where social media such as tweeter was used extensively.

Objective: The paper aims to provide a peek into the Iranian political landscape by quantitatively studying tweets collected during the 2013 Iranian presidential election with the help of Natural Language Processing (NLP) technologies. Specifically we are interested in the makeup of the online population who participated in the election one way or another, including their political orientation, their view on election participation, and their sentiment on individual presidential candidates.

Methods: We used Twitter search API to collect statuses containing hashtags and keywords relevant to Iran and Iran election. We collected data between May 21, 2013 and June 29, 2013 (Our data set cover tweets between May 14 and June 29). The data collection yielded in 3006528 tweets about Iran and election.

A language identifier was trained and used to identify the language of every tweet. For this study we focused only on the 460K Persian tweets that we identified. Classifiers were trained to automatically to identify the subject (political or non-political, election-related or otherwise) of a tweet, the candidates mentioned in a tweet and the sentiment towards them, the political orientation of the tweeting user and his/her view on election participation. In addition, a classifier also detects sarcasm in the tweet.

We manually annotated data to serve as the training and test data for the classifiers. We started with a pilot annotation, in which three annotators annotated the same 300 tweets. The pilot annotations were reviewed and adjustments were made to the guidelines before we annotated another 3,600 tweets - 1,200 tweets per annotators with 200 tweets overlapping between each pair of annotators. The trained classifiers were run on the 460K Persian tweets and quantitative analyses were conducted on the results.

Results: The results of this study show that the sentiment of Persian tweets were very positives towards the reformist candidates and very negative regarding the conservatives. Our findings reveal that the political environment of Persian Twitter is highly dominated by Reformist (pro-democracy) groups and confirms the findings of our earlier study on the community detection on the Twitter communication networks (mention and retweet networks) (Citation Removed). This suggests a difference between Persian Twitterverse and Blogosphere. Kelly and Etling (2008) in their study of Persian blogosphere suggested a more balanced environment, divided evenly between reformist and conservatives. Therefore, we can argue that different social media environments have different political landscapes.

Conclusions: The paper demonstrates NLP technologies, proven critical and effective in quantitative studies of political discourse in social media, lend us a way to examine social communication in a depth and at a scale that never existed before. We plan to expand our research to other languages and genres in the near future.

 

Refenreces:

Kelly, J., & Etling, B. (2008). Mapping Iran’s Online Public: Politics and Culture in the Persian Blogosphere. Berkman Center. Retrieved from http://cyber.law.harvard.edu/publications/2008/Mapping_Irans_Online_Public

 


Speakers
EK

Emad Khazraee

University of Pennsylvania
XM

Xiaoyi Ma

University of Pennsylvania


Saturday September 27, 2014 10:41 - 11:00
TRS 1-003 Ted Rogers School of Management

Attendees (4)