# train trigram language model, and compute <title,comment> pmi

    vocab.py            # static vocabulary

    ngram.py            # train trigram language model
                        # Kneser-Ney smoothing method is used.

    computePmi.py       # conditional probability is computed using "conditional corpus".
                        # simple smoothing method is used.

    computePmiUsLm.py   # conditional probability is given.
                        # Kneser-Ney smoothing method is used.

    averagePmi.py       # compute "average pmi" of each comment

    proc.sh             # the shell to run those programs, and does some processing.

    statistics/         # the directory to save trigram model and vocabulary.