Well, the documents being stored on FileNet were -mostly- forms, where the data was OCR'd out by a purpose-built industrial dedicated-machine installation/process, millions and millions of forms ...