Citations, IDs and Bates numbers
How to read a citation like DOC-0417 p.2 or VID-0012 08:41, how documents, pages and Bates numbers relate, and why every claim should carry one.
#Why citations matter
A finding you can't check is worse than no finding. Everything in Discovery Engine (search results, your AI agent's answers, highlights, the exhibit index) points back to an exact place in an exact document, so you or the attorney can open it and see it with your own eyes.
#The forms a citation takes
| Citation | Means |
|---|---|
DOC-0417 p.2 | Page 2 of document DOC-0417. |
EMAIL-0007 | The e-mail EMAIL-0007. |
EMAIL-0007 msg 3 | The third message in EMAIL-0007 (for a mailbox file holding several messages). |
SHEET-0033 row 118 | Row 118 of a spreadsheet. |
SHEET-0033 Ledger!row 118 | Row 118 of the sheet named Ledger in a workbook. |
AUD-0004 00:16 | The audio recording AUD-0004, at 16 seconds in. |
VID-0012 08:41 | What is said in video VID-0012 at 8 minutes 41 seconds. |
VID-0003 00:05 screen | What the video's screen shows at 5 seconds (text read from the picture). |
HFC-000123 | A Bates number (see below). |
In the app, a citation appears as a small chip. Click it to open the document at that place. Your agent prints the same strings, and can pass them straight back to the app's command-line tool, so a citation never has to be translated.
#Documents, files and pages
Four things are kept apart, and it helps to know which is which:
- The original file: what's on your disk, such as
wire-request-form-2019-03-03.pdf. - The document: what the app reads it as, with an ID like
DOC-0095. Usually one file is one document, but not always. An e-mail's attachments become documents of their own (each says which e-mail it was attached to), and a ZIP archive holds many. - The page or part: a page, paragraph, row, message, or stretch of a recording. This is what a citation points at.
- The Bates number: a stamp on a page in a formal production.
#Copies
If the same file appears twice, byte for byte, the app reads it once and counts the other as a duplicate. Documents with the same text (say, the same e-mail saved twice in different formats) are recognized as copies, so you and your agent don't read the same thing twice.
#Bates numbers
When a production is Bates-stamped, the app reads the stamp on each page and keeps it with that page.
- A document's Bates range (first to last) is shown at the top of the document viewer.
- A Bates number opens the page it is stamped on. Ask your agent about
HFC-000123, or click a link to it, and you land on that page. - A ten-page e-mail stamped
HFC-000920toHFC-000929is one e-mail on ten pages, not ten documents. The app never splits a document because its Bates number changes from page to page.
The app is careful about what it treats as a Bates number. It needs five digits or more, and it won't mistake its own IDs or a camera's file names (IMG_4471.jpg) for a stamp.
#Production load files
If the production came with load files (.dat and .opt), the app uses them to put page images together into documents. The load files themselves are listed on Progress as Production load files rather than read as evidence.
#How your AI agent cites
The prompt and guide the app gives your agent tell it to:
- cite every claim with the file's real name and the citation, exactly as printed (for example March statement.pdf, page 2:
DOC-0417 p.2); - read the passage before relying on it;
- end every answer with a Sources section: one line per document, with the file name linked to the original file, the citation linked to the document in the app, and the file's path in plain text.
When you check an answer, click the citation link (it opens the app at the passage). If your chat window won't open it, find the document by its ID on the Documents screen, or search for the quoted words.