The Party is Over Here: Structure and Content in the 2010 Election
In this work, we study the use of Twitter by House, Senate and gubernatorial candidates during the midterm (2010) elections in the U.S. Our data includes almost 700 candidates and over 690k documents that they produced and cited in the 3.5 years leading to the elections. We utilize graph and text mining techniques to analyze differences between Democrats, Republicans and Tea Party candidates, and suggest a novel use of language modeling for estimating content cohesiveness. Our findings show significant differences in the usage patterns of social media, and suggest conservative candidates used this medium more effectively, conveying a coherent message and maintaining a dense graph of connections. Despite the lack of party leadership, we find Tea Party members display both structural and language-based cohesiveness. Finally, we investigate the relation between network structure, content and election results by creating a proof-of-concept model that predicts candidate victory with an accuracy of 88.0%.
![Twitter and Politics14% of Internet users engaged in political activity via social media in 2008 => 22% in 2010 [Smith 2008, 2010].Political Polarization on Twitter ¨C [Conover et al. 2011]Characterizing Debate Performance via Aggregated Twitter Sentiment ¨C [Diakopoulos& Shamma2010]2Seal of Approval](https://image.slidesharecdn.com/icwsm-2011-avishay-livne-110719150754-phpapp02/85/The-Party-is-Over-Here-Structure-and-Content-in-the-2010-Election-2-320.jpg)
![Twitter in PoliticsPredicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment ¨C [Tumasjan et al. 2010]Hypothesis: Portion of tweets mentioning a party => portion of votes this party will get.Achieved MAE of 1.65%, traditional polls achieved 0.8%-1.48%.German federal elections 2009.In the same time a German octopus was used to predict the World Cup ¡®10 results...¡ correctly predicting 8/8 matchesPaul the Octopus3Our goal: Can we extract more information from Twitter?](https://image.slidesharecdn.com/icwsm-2011-avishay-livne-110719150754-phpapp02/85/The-Party-is-Over-Here-Structure-and-Content-in-the-2010-Election-3-320.jpg)
![Graph AnalysisDemocraticRepublicanTeaParty???????=???????????????????=????1?Supports previous studies [Adamic & Glance 2005]11](https://image.slidesharecdn.com/icwsm-2011-avishay-livne-110719150754-phpapp02/85/The-Party-is-Over-Here-Structure-and-Content-in-the-2010-Election-11-320.jpg)
The Party is Over Here: Structure and Content in the 2010 Election
- 1. The Party is Over Here: Structure and Content in the 2010 ElectionAvishay Livne, Matt Simmons, Eytan Adar, Lada AdamicUniversity of MichiganICWSM 20111
- 2. Twitter and Politics14% of Internet users engaged in political activity via social media in 2008 => 22% in 2010 [Smith 2008, 2010].Political Polarization on Twitter ¨C [Conover et al. 2011]Characterizing Debate Performance via Aggregated Twitter Sentiment ¨C [Diakopoulos& Shamma2010]2Seal of Approval
- 3. Twitter in PoliticsPredicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment ¨C [Tumasjan et al. 2010]Hypothesis: Portion of tweets mentioning a party => portion of votes this party will get.Achieved MAE of 1.65%, traditional polls achieved 0.8%-1.48%.German federal elections 2009.In the same time a German octopus was used to predict the World Cup ¡®10 results...¡ correctly predicting 8/8 matchesPaul the Octopus3Our goal: Can we extract more information from Twitter?
- 4. AgendaBackgroundDataNetwork analysisContent analysisPredicting election results4
- 5. Getting The Data5Unlike other studies we looked at candidates¡¯ accounts.Search: ¡°Harry Reid site:twitter.com¡±
- 6. 687 manually filtered users, 50% of the candidates.Data6*Tea Party ¨C political movement, supported Republicans.Gained wide attention in 2010.
- 7. Data460K tweets over 2 years + 233K URLs.4429 directed edges: A?B == user A follows user B7Health Care Reform Bill Passed
- 9. AgendaBackgroundDataNetwork analysisContent analysisPredicting election results9
- 10. Graph AnalysisEdge(A,B) = A follows BDemocratsRepublicansTeaParty10
- 11. Graph AnalysisDemocraticRepublicanTeaParty???????=???????????????????=????1?Supports previous studies [Adamic & Glance 2005]11
- 12. AgendaBackground & dataNetwork analysisContent analysisPredicting election results12
- 13. Language ModelingHow frequent each candidate/party used each term?13Extract distinguishing terms with KL-divergence
- 15. Party CohesivenessHow similar are pairs within each party15Similar Not similar
- 16. Language Similarity vs. Graph Distance16With retweetsWithout retweetsNot similar Similar The closer two candidates are the more similar their content is.This trend diminishes after 3 steps.
- 17. Latent Topics17Distribution of documents over topicsTopics1LDADocumentsDocuments2Average overparty¡¯s documentsDistribution of parties over topicsTopicsWhich topics are more affiliated with each partyCompare distributionsParties
- 18. Topics18
- 19. AgendaBackground & dataNetwork analysisContent analysisPredicting election results19
- 20. Predicting Election Results - Individual Features20ExternalNetworkUsageContentLogistic regression, 10-fold cross validation.
- 21. Predicting Election ResultsModels Comparison21Next, top model¡
- 22. 22Predicting Election ResultsTop model¡¯s coefficients
- 23. SummaryTwitter is prevalent in campaignsTea party more aggressive and Twitter-savvyRepublicans more aligned (network)Democrats less cohesive (content)Content similarity vs. network distanceNetwork and language improve prediction23
- 24. Future WorkMore extensive topic & sentiment analysisTemporal analysisIntegrating fundingIntegrating constituents (the crowd)24
- 25. ThanksThank you!Thanks to Abe Gong for helpful insights.More information in the paper.Contact info: avishay@umich.edu25