Generating Simulations of Motion Events from Verbal Descriptions