The following are the results you should get when computing cross entropy of the two given files using WITTEN BELL smoothing. Very slight differences are plausible depending on if you use the dummy words at the beginning of the text. If the effect of this is even visible, it will be quite small. bigrams (N=2) 19.1166583 0 396078 0.44894 1 736 0.40197 2 49 0.05352 3 17 0.02785 4 9 0.01966 5 5 0.01365 6 2 0.00655 9 1 0.00492 12 1 0.00655 13 1 0.00710 17 1 0.00928 0 395996 0.45200 1 806 0.40300 2 60 0.06000 3 23 0.03450 4 6 0.01200 5 5 0.01250 6 2 0.00600 12 1 0.00600 28 1 0.01400 trigrams (N=3) 28.7906387 0 250046044 0.48651 1 927 0.47176 2 19 0.01934 3 4 0.00611 4 4 0.00814 5 1 0.00254 11 1 0.00560 0 250045934 0.49306 1 1041 0.48150 2 20 0.01850 3 5 0.00694