How documents are read
Every file filed on a matter is read in the background. Reading extracts the text with the page it came from, which is what lets a finding in the analysis point at a page rather than at a document.
Until a document has been read it has no pages, and nothing in it can be cited.
The reading states
| State | What it means | What to do |
|---|---|---|
| Queued | Filed and waiting its turn. | Nothing. Pages appear once it has been read. |
| Reading | Being read now. | Nothing. Pages and citations follow when it finishes. |
| Stalled | Queued, and it has not moved since. | Nothing will pick it up on its own. Ask for it to be read again. |
| Unreadable | Reading did not finish. No pages, nothing citable. | Ask for it to be read again. If it keeps failing, the file is likely damaged or is a scan the parser cannot handle. |
| Not read | Nothing has read the file yet, so its length is unknown. | |
| No file | Recorded on the matter with nothing behind it. | Nothing to read. Upload the file if you have it. |
The number of documents currently being read is shown above the list.
Reading a document again
Read again from the document’s row menu, from its detail sheet, or on a selection of documents at once from the bulk bar.
Bytes already read under the current settings are re-used rather than re-parsed, so a re-read of a healthy document is cheap. It is worth doing when a document is stalled or unreadable, or after the file itself was replaced.
What reading does not do
Reading does not classify the document, decide its date, or judge it. It produces text and page boundaries. Everything else — the type, the origin, the document date, the sensitivity flags — is yours, and the analysis reads those back. A document filed as a deposition is read as one.
Where a browser cannot draw the file type at all, the viewer falls back to showing what the parser read out of it. If it has not been read yet, there is nothing to show — which is what Nothing to show yet in the viewer means.
Duplicate detection
As documents are read, their text is compared against what is already on the matter, and near-duplicates are proposed for a decision. Nothing is merged either way. See Duplicates and relations.