ngrams.py

#!/usr/bin/env python

import collections
import sys

def tokens(f):
    for line in f:
        for t in line.split():
            yield t

def get_ngrams(stream, n):
    stream = iter(stream)
    window = collections.deque(maxlen=n)

    while len(window) < n:
        window.append(stream.next())

    for t in stream:
        yield tuple(window)
        window.append(t)

def with_counts(it):
    seen = {}
    for x in it:
        seen[x] = seen.get(x, 0) + 1
    return ((y, x) for x, y in seen.iteritems())

def main():
    ngrams = get_ngrams(tokens(sys.stdin), 2)
    counts = sorted(with_counts(ngrams), reverse=True)
    for x, y in counts:
        print x, ' '.join(y)

if __name__ == '__main__':
    main()