Let’s talk about transcription in scholarly editions – those collections of “papers” by or about historical or literary figures, published for readers and researchers. There are basically three ways edition projects transcribe papers. The first is the old-school way with staff members doing it manually. The second is a group effort, which can be done through crowdsourcing or smaller research teams. And the third is using machine techniques like HTR or OCR to do it automatically.
Now, the big question with machine transcription is, how accurate is it? If it's 98% accurate, is that good enough for a scholarly edition? If so, it’s a cost-effective way to produce an edition. But what about the other benefits of human transcription? When people transcribe, they don't just create a digital copy of the text, they also learn a lot about the content, the people mentioned, and all sorts of other interesting things. So, if we rely solely on machines, we could miss out on all of that extra knowledge.
Plus, imagine a world where everything is automated, and we end up with digital scholarly editions made up of texts that no one has ever even read. That's a bit of a scary thought to me!
So, while machines can definitely help us out with transcription, we shouldn't forget about the importance of human involvement. We need to find a good balance between automation and old-fashioned human effort, so that we can create digital scholarly editions that are accurate, insightful, and affordable.