A large proportion of human noncoding DNA had been known to have no biological function. However, unprecedented technical advances have started to convert unannotated noncoding DNA into highly annotated functional regions. In this paper, the frequency of n-grams of regional DNA sequences from fifteen chromatin states of Broad Histone Track are thoroughly analyzed, applying biological language modelling to n-grams. It has been shown that a few particular n-grams are found in abundance in one chromatin state but occurring very rarely in other states, thereby serving as chromatin state signatures. We discuss the significance of the patterns found, as well as their potential use in the special statistical models of nucleotide sequences necessary for developing algorithms for the computational analysis of functional units in noncoding DNA regions.