IRC NOMMER - Sentiment analysis of #python and #ruby-lang

What?

IRC NOMMER eats IRC messages, sends them off to Chatterbox.co's sentiment analysis API and then adds the message's score to an average for the IRC channel. The result gets sent to Cosm.com where it is logged, and makes these graphs.

Why?

Because a RoR developer (@donkeyscience) got ragey at the OSS community's "underhelpful" interactions on IRC. A Django developer (@stringfellow) quipped that the Django community was ace. And they felt like doing SCIENCE!!!

How?

While this isn't quite science, it is the result of a hacky afternoon with cosm, chatterbox.co and IRC. It's a bit of fun. We pitched #ruby-lang against #python (and some other languages)

Methodologyish

The sentiment of each message is acquired from the chatterbox API. Then, the value is put in a fixed-sized array (20). The average value is calculated on this array, and that value is sent to cosm. When a message hits any of the rooms, for any room that has had no activity for greater than 5 minutes, a zero-value is pushed to the array, the mean calculated, and the new value sent to cosm, the 5 minute timeout is then reset (so that quiet rooms tend toward neutral).

Privacy

No usernames are sent to chatterbox.co (except where they appear in a message), or stored anywhere.

Notes

The sentiment values are between -1 (very negative) and 1 (very positive).
Due to some pretty stupid maths, the values before 2013-02-10T01:25:00 were total bullshit. Probably.
Sometimes the graph goes dotty cos no one has said anything, and in at least one case, because IRCNOMMER got kicked :(
If IRCNOMMER dies, or gets re-started, the running average is destroyed at the moment.
IRCNOMMER tweets once every half hour to say who's "winning".