Thanks Nick for your gesture. I will certainly be interested in trying out your cached tweets, but its usefulness will be limited to those who follow the cached tweets' authors.
About sharing, i don't intend to publish in journals or conferences as i can't afford the costs, but will definitely share any insights from my experiments. The idea is simple : Current twitter interface either lacks groupings or some apps provide ad-hoc groupings like TweetDeck. But I would like to develop an app that would cluster semantically homogeneous tweets and present the result to a user in a nice GUI. The app would be free for the users and if it works fine, at a later stage could be supported by some sponsors. But it all depends on Twitter if it would allow a one-time download of a user's recent past 100,000 or so friends_timeline to build the model. It could change the twitter client game completely as we dive deeper into the meanings of the tweets instead of the keyword based or author based groupings.