Project 1: Attack on State of Art lexical resource journey
This was the first project under the belt of sentiment analysis. It aimed validating the present lexical resource(you name it :D) for its correctness and definitions. And the result suggest that "ummm...ohhh....okkk" type. More coming on after the report gets validated. More on this after Jan 15,2010.
The verdict is out. The resource we did our experiment is SentiWordNet-> for more details chip in a mail, I would be happy to share my result.
Project 2: Boost up accuracy of a SA system
What are the good features for SA classification- or did we find them before, but just forgot them. All the experiments related to this endeavor aim at boosting the Pang & Lee accuracy to never before accuracy and i must say results are promising. This is a joint venture with Debraj Manna (M.Tech 2nd yr CSE,IITB) and so far its quite on track.
Project 3: Finding core concepts
Status: in queue
What are the core concept that kid learns when starts his life - this is closely related to my PhD work. I have devised a way to empirically get the lower bound on the inane concepts that kids should have in order to understand deeper semantics this complex world has to offer for him. The proof is geometry based
Project 4: Cross domain sentiment analysis
In this work, the issues related to cross domain sentiment analysis are addressed. We are experimenting with various techniques to create system which can automatically tag the documents in other domains