OCR is mostly good enough. Problem here is we have 76 pages that we need to be read perfectly, with a low fidelity input

We also have very little in the way of error correction, since it’s mostly not human readable