The project is investigating methods to extract human-marked data from document forms such as surveys and tests. They can read questions, multiple-choice exam papers, and grade.