Numbers for this chart will change as we finish collecting data.
We're using a far larger dataset now: 110,000+ tweets. We're also using Google's Prediction API for more accurate classification. However, that slows us down a bit, so numbers will update over time.
'Neutral' is for any tweets that talk about the shutdown but don't attempt to assign blame.
Portent! An internet marketing agency and technology company in Seattle, WA.
We collect tweets using Topsy, then use machine learning (Bayesian class-ification, if you must know) to figure out if the tweet is referring to the Democrats or Republicans in the context of the shutdown.
Data from Topsy. Analysis tools built using Python and TextBlob. Pretty graph built using TufteGraph.js.
Sometimes, the tweet counts for one party or the other may get smaller. The algorithm 'learns' over time. That may mean it adjusts some numbers downward.