Addressing the limitations of current document AI approaches by focusing on few-shot relational learning in visually-rich documents.
