Networks and language in the 2010 election
|Conference paper (help)|
|Networks and language in the 2010 election|
|Authors:||Avishay Livne, Matthews P. Simmons, W. Abraham Gong, Eytan Adar, Lada A. Adamic|
|Citation:||missing booktitle in Conference Proceedings at OpenSIUC : 15. 2011|
|Meeting:||4th Annual Political Networks Conference|
|Web:||DuckDuckGo Bing Google Yahoo! — Google PDF|
|Article:||Google Scholar PubMed|
|Restricted:||DTU Digital Library|
Networks and language in the 2010 election describes a text mining and network analysis of Twitter data generated by United States House, Senate and gubernatorial candidates for the midterm (2010) election.
The analysis also included text sentiment analysis.
The paper is very similar to The party is over here: structure and content in the 2010 election and seems to present the same results.
The Twitter data consisted of 460'038 tweets from 2007 to 2010 from 687 candidates. 339 democrats and 348 republicans (including 95 identified as Tea Party candidates).
The Twitter network between the candidate was also crawled. 4'429 edges between candidates were found.
The researchers also downloaded the candidate homepages as well as web pages linked from the tweets. The found 132'376 distinct web pages.
- Network analysis. Graph visualization, density, in-degree.
- Ordinary statistics, e.g., "tweets per day"
- Topic mining with Latent Dirichlet analysis using GibbsLDA++
- The sentiment analysis was performed with the AFINN word list (A new ANEW: evaluation of a word list for sentiment analysis in microblogs).
- "Content cohesiveness"
- Candidate election results prediction based on features such as closeness, authority, PageRank, in/out-degree, incumbency and KL-distance to party, sentiment, tweets statistics
The paper presently states "Please do not quote from or cite this paper without the authors' permission". However, The party is over here: structure and content in the 2010 election seems not to have such reservations.