Soft launch: Soldier biographies and research notes vary in completeness. Prefer class-roster, NARA, newspapers, and other cited sources for contested facts — treat Research notes as leads, not settled history. Anyone can correct a page; contact us with questions.

Contribute documents: Difference between revisions

From Ritchie Boys Wiki
Jump to navigation Jump to search
Simplify How it works for public contributors
Distinguish soldier OCR from Training materials
 
(3 intermediate revisions by the same user not shown)
Line 2: Line 2:
<div class="ocr-inbox-slot"><p>Loading document OCR tools… (requires login and the OCR review service).</p></div>
<div class="ocr-inbox-slot"><p>Loading document OCR tools… (requires login and the OCR review service).</p></div>


Site-wide inbox for document OCR. Upload a scan, confirm '''who''' it is and '''which soldier page''' would change, then Apply.
Site-wide inbox to '''enrich soldier profiles from document scans''' — forms, certificates, clippings, letters, and team rosters. Upload a scan, confirm '''who''' it is and '''which soldier page''' would change, review proposed infobox facts and a short biography draft, then Apply.


== How it works ==
== How it works ==
# '''Log in''' (same as photo uploads).
# '''Log in''' (same as photo uploads).
# Upload a document (photo or PDF of a record, form, list, or similar). OCR runs on the server; nothing is written to a soldier page yet.
# Upload a document (photo or PDF). Printed text uses the free local OCR engine. Check '''Handwriting / hard scan''' for cursive or faint pages — the service still runs a free print OCR first and only calls the paid LLM vision model when that preliminary read looks weak.
# '''Confirm''' who the document is about. If several people appear on one scan, you get a table: each person, the matched wiki page (when found), and what would be written on that profile. Names, ranks, and service numbers from the scan are matching cues — they are not applied until you approve.
# '''Confirm''' who the document is about. Review proposes '''infobox''' fields and a '''biography draft''' woven from clear facts — not a raw OCR dump. Nothing is written until Apply.
# For '''multi-person IPW / team rosters''', you get a table: each person, the matched wiki page, and what would be written (usually Team membership). Names, ranks, and ASNs are matching cues until you approve.
# Check '''Soldier pages that will be edited''' at the top. Use '''Preview Apply''' to see planned changes without writing. Then '''Apply selected''' (requires <code>ocr-reviewer</code> or admin), or '''Discard''' the job.
# Check '''Soldier pages that will be edited''' at the top. Use '''Preview Apply''' to see planned changes without writing. Then '''Apply selected''' (requires <code>ocr-reviewer</code> or admin), or '''Discard''' the job.
# Optional controls (raw OCR text, gallery attach, overwrite existing fields) are under '''Advanced'''.
# Optional controls (raw OCR text, document type, gallery attach, overwrite, handwriting re-OCR) are under '''Advanced'''.


You can also start from a soldier page → Gallery → '''OCR a document''' (that soldier is pre-selected).
You can also start from a soldier page → Gallery → '''OCR a document to enrich this profile''' (that soldier is pre-selected).


== Privacy ==
== Privacy ==
Line 20: Line 21:
== For reviewers ==
== For reviewers ==
Administrators can grant the '''ocr-reviewer''' group. Other logged-in uploads stay in the queue until a reviewer applies or discards them.
Administrators can grant the '''ocr-reviewer''' group. Other logged-in uploads stay in the queue until a reviewer applies or discards them.
See also [[Facebook OCR review]] (editors-only queue of Facebook-sourced document OCR candidates; review aid only — prefer this page for new uploads). Camp-wide training manuals belong on [[Training materials]], not here.


[[Category:Ritchie Boys Wiki]]
[[Category:Ritchie Boys Wiki]]

Latest revision as of 02:28, 15 September 2026

Loading document OCR tools… (requires login and the OCR review service).

Site-wide inbox to enrich soldier profiles from document scans — forms, certificates, clippings, letters, and team rosters. Upload a scan, confirm who it is and which soldier page would change, review proposed infobox facts and a short biography draft, then Apply.

How it works[edit | edit source]

  1. Log in (same as photo uploads).
  2. Upload a document (photo or PDF). Printed text uses the free local OCR engine. Check Handwriting / hard scan for cursive or faint pages — the service still runs a free print OCR first and only calls the paid LLM vision model when that preliminary read looks weak.
  3. Confirm who the document is about. Review proposes infobox fields and a biography draft woven from clear facts — not a raw OCR dump. Nothing is written until Apply.
  4. For multi-person IPW / team rosters, you get a table: each person, the matched wiki page, and what would be written (usually Team membership). Names, ranks, and ASNs are matching cues until you approve.
  5. Check Soldier pages that will be edited at the top. Use Preview Apply to see planned changes without writing. Then Apply selected (requires ocr-reviewer or admin), or Discard the job.
  6. Optional controls (raw OCR text, document type, gallery attach, overwrite, handwriting re-OCR) are under Advanced.

You can also start from a soldier page → Gallery → OCR a document to enrich this profile (that soldier is pre-selected).

Privacy[edit | edit source]

  • Do not upload Social Security numbers or private contact details for living people.
  • Staging files are not published until Apply (Gallery attach is optional; off by default for multi-person uploads).
  • See About the data and Contribute.

For reviewers[edit | edit source]

Administrators can grant the ocr-reviewer group. Other logged-in uploads stay in the queue until a reviewer applies or discards them.

See also Facebook OCR review (editors-only queue of Facebook-sourced document OCR candidates; review aid only — prefer this page for new uploads). Camp-wide training manuals belong on Training materials, not here.