That's pretty cool. If one were inclined to take it to the next level of absurd awesomeness, one could employ machine learning: 1. Feed all the examples from

That's pretty cool. If one were inclined to take it to the next level of absurd awesomeness, one could employ machine learning: 1. Feed all the examples from the Gregg dictionary into a neural network. 2. Generate all the possible positionings for a word. 3. Have the trained NN pick the best one. That said, I think doing the opposite conversion, from Gregg to English, would be marginally more useful (and a lot harder, I assume).