In the weeks and months leading up to elections, news outlets conduct poll after poll and jostle to predict who will win. What if they had predictions updated minute by minute, based on an even larger audience?
This is the idea behind ElectionGauge, a DC-based startup that analyzes Twitter response to events. Predicting election results is their first test of an algorithm that could analyze the state of everything from the economy to sports.
Cofounder Maksim Tsvetovat, a professor at George Mason University and coauthor of Social Network Analysis for Startups, has 12 years experience in social network analysis and social media.
“I was looking for an interesting project that would help me tie geo-spatial analysis, linguistic analysis, and social network analysis – and, serendipitously, it happened to be the election year. Scientifically, we had an opportunity to gather a large dataset and learn new and interesting things about the world,” he says.
Founded in September 2011, ElectionGauge predicts election results by matching the speech of candidates to the tweets of users. For example, “repeal Obamacare” might identify you as Tea Partier, while “legalize marijuana” puts you in Ron Paul’s camp. I imagine that political hashtags would go into the mix, too. (ElectionGauge analyzes only American Twitter users but can’t filter out those under 18, so that may affect results a bit.)
Tsvetovat says their algorithm has already predicted a few primaries, like Gingrich’s win in South Carolina. It missed the recently Santorum sweep due to a bug, which has since been fixed.
ElectionGauge hopes to go beyond Twitter buzz metrics (counting mentions) or sentiment analysis (guessing emotions) by bringing to bear linguistics and related fields. According to Tsvetovat, the project is non-commercial and they hope to see their data used by news outlets.
Learn more from the team at our Tech Cocktail DC Winter Mixer tonight.