I'm mainly using Ancestry to build my tree, accessing the many scanned documents they have. However, sometimes the scans are very difficult to read and additional details haven't been added to the metadata (for e.g., occupations listed on marriage certificates). More worryingly, I think I found an incorrect date. I'm 90% certain that the handwritten date is Oct 13 (on the previous page is a Sept 19, and the month looks different on the two pages, and Sept 13 coming after Sept 19 seems unlikely. To be clear, this isn't a fault with Ancestry, but an issue with anything that has relied on computer OCR and human volunteers (or lowly paid civil servants).
Anyway, it gave me an idea, but I wonder if it already exists. Does anyone know of a human-powered OCR project where users can upload images and ask volunteers to enter the text that they see? Google has run such things as game for machine learning, so it wouldn't surprise me if someone set up something like this, and I figure genealogists would be ideally placed to know the answer!
If no one is aware of such a website, would you find such a thing useful? (I know I would - I'm relying on friends and family to give me their opinions on words).