[pdurrant]

Unfortunately, there's no good automated way, as PDFs are mostly just characters positioned on a page, with no sense of flow.

Perhaps your best bet would be trying an OCR program on them, as if they were hard copy scanned in.