My solution to the Quora Classifier problem in Codesprint 2 (Interviewstreet)
It's basically a Naive Bayes classifier assuming the prior to be equally probable
First step is to classify whether nodes are discrete or continuous. Continuous nodes are assumed to have a Normal distribution.
Then training and classification is done. It's quite a hack! :)

To run: python file2stdin.py | python classify.py

Problem definition can be found here (newer version, but same problem):
http://linode.interviewstreet.com/CodesprintQuora-AnswerClassifierRedux.pdf