Redaction Analysis
2.6 million redaction bar locations mapped by spatial analysis across all EFTA documents. The <code>ocr_text</code> column contains text the scanner produced when its OCR engine tried to read the black bars — ~98% is garbled noise, not recovered hidden content. The EFTA documents use baked-in image redactions (not removable PDF overlays). Also includes 107,000 automatically extracted entities (names, dates, amounts).
← Search all databasesPer-document redaction summary. <code>text_near_bar_count</code> = bars where OCR text was detected nearby (mostly noise).
Names, phone numbers, amounts, emails, and organizations extracted from document text.
Detected redaction bars with spatial coordinates. <code>redaction_type</code>: text_near_bar (OCR text found near bar), proper_redaction (no text detected), white_rectangle. The <code>ocr_text</code> column is scanner noise from the OCR engine attempting to read black bars — not recovered hidden text.