Machine Learning

I have been working on a document that a client converted from PDF to Word using OCR. I am discovering that OCR is often wrong in predictable ways. Every instance of decr, an abbreviation, is now deer, and every Iuris, which is not English, is now Juris.

It is not news that machines still need us to fix their mistakes. Notwithstanding, I’m really grateful for that OCR: it's much faster to correct it than to fix the typos I would have made if I'd had to type a thirty-two page document. Similarly, I'm not opposed to clients using Grammarly or ChapGPT—but please don’t think they wholly replace a copy editor, or that they are perfect just because they are machines.