Table 1.

String counts in the Brown corpus. Component trigrams for the string “_ing_” are common in the corpus, which results in a high probability in the trigram model, despite the exact string rarely occurring.

String	Count	Probability
__i	68440	0.07
_in	33783	0.49
ing	30454	0.34
ng_	30035	0.78
_ing_	1	~0