Every so often, Zooniverse has a project of special interest to the iNaturalist community. Their latest is designed to train a machine learning algo to accurately transcribe notes on herbarium specimens. This is supposed to help save a lot of time in the specimen digitization process.
This is interesting. Is it entirely automated, or do you think it would benefit by knowledgeable people also manually reading over/checking the translations? In some cases, a person would be able to error correct possible mistakes so would be useful as a final review step, unless that’s already being done.
The whole project involves humans identifying text so a program can learn how. There are probably places where it is difficult to decipher the text, but Zooniverse is pretty good about providing “I don’t know” buttons so volunteers don’t get forced to make bad calls.