I used a “chart parser” architecture to cope with the highly ambiguous parse trees involved. How did I pick the one that was best? I thought - grade them all on how complex they were and pick the simplest possible. It worked pretty well. Later, the statistical analysis of language crowd came to the same conclusion. It turns out that Dr. Chomsky and his friends had spent a lot of time looking for a complex explanation for something that wasn’t.

I have always wanted to do a visualization of a parse tree intended to aid humans in figuring out foreign languages. Draw the diagram with labels on the branches and display lexicon entries for the words. Its like what you would do to translate a language when all you had was a dictionary - and let the human mind grasp the higher structure from the boost provided by having the lexical definitions brought readily to hand.