Site changelog: Difference between revisions

Line 12:

=== 26 August 2026 ===

* '''Document OCR + review''' — Logged-in users can upload a document from a soldier page or from [[Contribute documents]]; the site OCRs it and shows proposed facts for human review. Nothing is written to profiles until '''Apply''' (requires the <code>ocr-reviewer</code> group or an admin). Staging files are not published automatically. Docker service: <code>ocr-review</code> on port 8090.

* '''Facebook photo OCR (pilot → full review)''' — OCR pipeline with confidence % for document-like Facebook gallery images; pilot facts applied to Atchley, Hauboldt, and Hauser. Full-run review checklist (numbered): `import/facebook_posts/enrichments/FB_OCR_REVIEW.md` (+ `.json` / `.csv`). Re-run: `python scripts/fb_photo_ocr_pilot.py --all`. Incorporate by telling the agent item numbers or soldier names.

* '''Facebook photo OCR (pilot)''' — Script to OCR document-like Facebook gallery images, extract dates/ranks/ASN/class cues, and write a review report (no auto-edits to soldier pages). Pilot: `python scripts/fb_photo_ocr_pilot.py --pilot`. Reports: `import/facebook_posts/enrichments/FB_OCR_PILOT_REPORT.md`, `FB_OCR_PILOT.json`. Requires `pip install rapidocr-onnxruntime pillow`.

* '''About the data (disclaimer)''' — Expanded how site information is compiled: Dan Gross research files, NARA and census merges, Facebook imports, family and contributor submissions, AI-assisted summaries and cleanup, and open wiki editing—with explicit limits on accuracy. Class roster / museum reproduction language kept in [[Template:Class data disclaimer]].

* '''Facebook biography curation (batch 2)''' — Hand-curated biographies for the remaining ~26 Facebook-enriched soldier pages that still had roster dumps, birthday-post voice, gallery stubs, or cross-post fragments (IPW Team 34 roster, Lewy/Bergman family posts, etc.). Re-run: `python scripts/apply_remaining_bio_fixes.py`. Report: `import/facebook_posts/enrichments/BIO_REMAINING_FIX_REPORT.md`.

@@ Line 12: / Line 12: @@
 === 26 August 2026 ===
+* '''Document OCR + review''' — Logged-in users can upload a document from a soldier page or from [[Contribute documents]]; the site OCRs it and shows proposed facts for human review. Nothing is written to profiles until '''Apply''' (requires the <code>ocr-reviewer</code> group or an admin). Staging files are not published automatically. Docker service: <code>ocr-review</code> on port 8090.
+* '''Facebook photo OCR (pilot → full review)''' — OCR pipeline with confidence % for document-like Facebook gallery images; pilot facts applied to Atchley, Hauboldt, and Hauser. Full-run review checklist (numbered): `import/facebook_posts/enrichments/FB_OCR_REVIEW.md` (+ `.json` / `.csv`). Re-run: `python scripts/fb_photo_ocr_pilot.py --all`. Incorporate by telling the agent item numbers or soldier names.
+* '''Facebook photo OCR (pilot)''' — Script to OCR document-like Facebook gallery images, extract dates/ranks/ASN/class cues, and write a review report (no auto-edits to soldier pages). Pilot: `python scripts/fb_photo_ocr_pilot.py --pilot`. Reports: `import/facebook_posts/enrichments/FB_OCR_PILOT_REPORT.md`, `FB_OCR_PILOT.json`. Requires `pip install rapidocr-onnxruntime pillow`.
 * '''About the data (disclaimer)''' — Expanded how site information is compiled: Dan Gross research files, NARA and census merges, Facebook imports, family and contributor submissions, AI-assisted summaries and cleanup, and open wiki editing—with explicit limits on accuracy. Class roster / museum reproduction language kept in [[Template:Class data disclaimer]].
 * '''Facebook biography curation (batch 2)''' — Hand-curated biographies for the remaining ~26 Facebook-enriched soldier pages that still had roster dumps, birthday-post voice, gallery stubs, or cross-post fragments (IPW Team 34 roster, Lewy/Bergman family posts, etc.). Re-run: `python scripts/apply_remaining_bio_fixes.py`. Report: `import/facebook_posts/enrichments/BIO_REMAINING_FIX_REPORT.md`.