Since Twitter is great for data mining, i hacked together a prototype, also since the weather was great this weekend it's really a hack that needs some (well a lot) of rework, probably need to dive into python multithreading, or reimplent the whole stuff in another programming language, would be a cool project to dive into go once.

The Idea is to use tweepy's streaming API to filter tweets for up to 512 words stored in a wordlist. 
Then whenever a tweet is written, the coresponding LED gets a bit brighter, until it overruns an start again, so you'll end up with LED blinking faster on often used keywords and slower on less often. (while writing this description: pulsing would be cool too)

Sorry no video right now, because I really just finished the first code yesterday night, and need to think of 512 keywords first.