My deleted tweet visualizations have been through several concepts that were practice for the real thing. I experimented with Parts-of-speech tagging, Automatic text classification and other tools in an effort to dig deeper into the tweets that people take back.
Below are some prototypes that inspired the final product.
Text categorization and tweet lifetime (how long it was public before being deleted):
User location, platform (red is desktop, yellow is mobile, blue is text message) and number of followers (size of circle)
Visualizing the time of day tweets were sent and deleted at:
Investigating sentence structure through Parts-of-speech tagging. Accuracy wasn’t very high due to the limited content to work with.