I guess I don't see what "automata theory" as such has to do with NLP, New Code or otherwise. As an *applied* mathematician, I was very excited by automata theory, formal logic, theorem proving by machine, formal semantics of programming languages and other related disciplines, back about the time John and Richard were inventing / discovering NLP. But all of the dreams I had of *practical applications* of these studies fell by the wayside. I now view them as a dry hole, although I'm sure my friends in cryptography have uses for them. :) I find the models of John Holland's Complex Adaptive Systems much more inspiring, for example, as I do the practical techniques of artificial intelligence, artificial life, genetic programming, fuzzy logic, simulated annealing, etc. IMHO, NLP is about what works -- these tools work. Once you get the basics of automata theory down, it's pretty much a bunch of theorems about what doesn't work. It's impossible to solve the halting problem, this or that problem is either unsolvable or requires exponential time to solve, you can't write an optimizing compiler for more than three registers, etc. I like having a bag of tools that give me usable solutions to real-world problems, not a bunch of theorems about what I can't do.