Of course Julia can work fast with Array, I know. But in natural language processing or text analyzing, we often count word frequency and create dictionary. We usually store word frequency in kind-a Dict and we always cut off non-frequent words (its frequency are under threshold) to exclude noisy words. So I want remove keys which values follow some condition.
Finally, I found John Myles White's implementation creating n-gram. So, I will refer this. https://github.com/johnmyleswhite/TextAnalysis.jl/blob/master/src/ngramizer.jl