What happens when we ask a machine to look at human communication, without direction? TweetMind grabs tweets matching a certain search term, and attempts to divine their qualities by analyzing the character frequency in the tweets. The viewer observes both the incoming (mostly) human tweets and TweetMind’s progress in constructing its own tweet from the others. Initially, the TweetMind tweet is small, but it evolves as new data is observed, eventually settling on its understanding of the meta-tweet.
TweetMind currently processes tweets matching a search of “news”. This was chosen to ensure a high amount of churn in the tweet search results. After experimenting with the ofxSelfOrganizingMap addon, I can see that this particular machine learning algorithm is not well-suited to building similar messages, but it does provide a different window into the Twitterverse.
TweetMind was developed using OpenFrameworks and relies upon the ofxTwitter, ofxHttpUtils, ofxXmlSettings, and ofxSelfOrganizingMap addons. It has been tested on Arch Linux 64-bit. You can check out the source on Github.