It could be possible to scan the PDF and automatically find the stat blocks within it, making it easier to extract them.