Triage for large documents

What Triage is for

A 900-page records request, a long discovery production, a multi-chapter case file — these are the kinds of documents where a per-match review on every page is the wrong tool. Triage sorts pages into seven buckets by their content profile, so you can keep the pages you actually need to share and drop the rest before redaction starts.

Triage opens automatically on documents over about 50 pages. You can turn it off in Settings if you always want the standard per-page review, even on long files.

How to triage a long document

1. Lex Cloak opens the Triage view with seven buckets across the top: high-fidelity, pre-redacted, lower-fidelity, cover and metadata, numeric, boilerplate, and sparse. Each bucket shows the pages it classified, with a count.
2. Click into a bucket to see its pages in a grid. Click any page in the grid to open the lightbox for a full-size view; use the arrow keys to step through pages without leaving the lightbox.
3. For each page or bucket, choose Keep to send it to the normal redaction review, or Drop to leave it out of the saved file entirely.
4. The toolbar pill shows your overall progress: how many pages are decided and how many still need a call.
5. When you are done, the kept pages go through the normal review loop. The dropped pages are simply absent from the export and your original document is untouched.

The seven buckets and what they mean

• High-fidelity. Dense pages of original content. These are usually the pages you keep and review carefully.
• Pre-redacted. Pages that already carry visible redactions. Worth a glance to confirm the redactions match your standard.
• Lower-fidelity. Pages with weaker text recognition, often scans of poor quality. Drawing or whole-page redaction may be faster than reviewing match-by-match.
• Cover and metadata. Title pages, signatories, transmittal letters. Often safe to keep or drop wholesale depending on the audience.
• Numeric. Tables, ledgers, lab values. These often carry identifiers in the surrounding text that detection finds, but the numbers themselves usually do not need redaction.
• Boilerplate. Repeated language: instructions, certifications, oaths. Often safe to keep without page-by-page review.
• Sparse. Mostly empty pages. Usually a drop unless the small amount of content matters to your purpose.

The app seems frozen on a large document

On a long document, Lex Cloak does a heavy round of OCR and name analysis before per-page progress appears, and the status reads Analyzing names. Give it a minute or two; it is reading the whole document first. After that, Triage opens.

For more on common stuck-or-error cases, see common errors.