I ran the ant script on a bunch of volumes one by one, and then scripted it to run on all the remaining ones, which it's steadily doing (between one and two hours for each, I think). I've edited the "Old issues" page to reflect what we're doing and uploaded the first six quick-and-dirty PDFs. This morning KS did a quick count of errors by word over five pages, and we're hitting better than 99% accuracy on those pages, but there are other places where it's poorer and that doesn't take account of italicization and bold (more of that below).

KS has already corrected the text for volume 14 (remarkably quick), and is now doing a second pass looking at italics and bold. It turns out that Tesseract encodes that feature at the word level, by adding em or strong elements inside the word. I've created Oxygen project actions to italicize or bold an element (or undo same) as an experiment with Author Mode customization, and that works, but it's not ideal because it's tedious to do a whole line. I'm now trying to use a custom XSLT action to transform a selected fragment to add or remove em elements to every word, but I've hit a snag with that and had to post to the Oxygen forum for help. If it proves possible to get that working, I can imagine a lot of uses for it.