I’ve managed to get all of the various components wired in the right locations. Data is now flowing from Twitter into RabbitMQ, through a pair of workers, past a sentiment analysis API, and through to an in-browser visualization. The results are gorgeous!
The animation above is just a preview of what the data looks like. Because I don’t have access to the whole Twitter firehose, I’m limited to just a subset of the sample data (roughly 1% of total traffic). I then further limit myself to just English tweets so the sentiment analysis API works appropriately. Then, because we’re looking now at only 1% of 20% of the data, I limit even further with a geographic bounding box around North America.
Not every tweet has geolocation data, so I can only plot the ones that do. This is a fun chart showing relative, geographical sentiment, but unfortunately isn’t enough data at a high enough bitrate to really satisfy my itch.
I might try to take this a bit further and do some keyword latency/frequency mapping with sentiment as well – think of a live tag cloud where terms grow with usage – so I won’t be bound to geolocation …