This project investigates methods of using natural language generation to automatically produce disambiguating paraphrases of structurally ambiguous sentences in order to enable crowd-sourcing of data for parser training and domain adaptation.