On my (underpowered) laptop, this program initializes the dictionary-trie from my ~0.5MB unix dictionary file in 0.18 seconds and it checks/corrects the entire text of Huckleberry Finn (~0.5MB as well — and a lot of words that need correction) in a little less than a second.

This basic efficiency should also extend to most changes made to this method to make it more sophisticated (e.g.: sorting on likelihood of a word in context).