DOJ DataSet 12 · four files
Help us read the cipher pages.
Four documents from the Justice Department’s DataSet 12 release are mirrored here page by page, 101 pages in all, and 39 of them are hand-printed in a cipher. The enciphered pages are the ones that need reading, and machine OCR failed on every one of them, which is why they are still unread: search cannot reach a page nobody has turned into characters. So the work is done by hand, in the open. You type what you see, choose how you think it should be read, and record it. When a second member arrives independently at the same thing, the page moves on. Nothing here is scored and nothing is timed — reading a government record carefully is slow, and it is meant to be.
The rest of the pages are not cipher and the workbench does not offer them: typeset poems, magazine cuttings, photographs, some pages of ordinary handwriting, one blank leaf. Read together, the four documents are somebody’s scrapbook — poems about captivity, collages about a child. That is the record these enciphered pages sit inside, and it is worth knowing before you start.
Reading, the transforms and the analysers need no account and never will — open any page and use everything on it. Putting a reading on the record is the one thing that does: it needs a free account and a community handle, because the work is credited to you and a second member has to be able to agree with it independently. Worth doing before you start on a page rather than after.
We publish the page image and nothing else. There is no transcript of these pages on this site, from us or from anyone, because the first text anyone sees becomes the text everyone corrects instead of reading the page — and our mistakes would then turn up in every submission after it. Every page carries the URL of DOJ’s own copy and the fingerprint of the PDF we rendered it from, so you can check our mirror against the original before you spend an hour on it.
“Transcribed” means two members typed the same characters. “Solved” means two members also arrived at the same reading of them, and the server was able to re-derive that reading from what they typed. Neither is a claim about what the page means — only that the same work has been done twice, by two people, and can be repeated by you.