{
 "title": "OCR recall from a full-page screenshot PDF",
 "measured": "2026-08-01",
 "source_document": "English Wikipedia article 'PDF', captured as one continuous PDF (700 x 11008 pt, no text layer)",
 "ground_truth": {
  "source": "Wikipedia REST API, same day",
  "words": 10285,
  "characters": 64319
 },
 "tools": {
  "render": "poppler-utils pdftoppm",
  "ocr": "Tesseract 5",
  "flags": "--psm 6 -l eng, OMP_THREAD_LIMIT=1"
 },
 "full_page_run": {
  "render_dpi": 150,
  "render_seconds": 29.98,
  "image_pixels": "1459x22934",
  "image_bytes": 22105092,
  "ocr_seconds": 90.23,
  "ocr_words": 10343,
  "normalised": {
   "ground_truth_words": 8122,
   "ground_truth_distinct": 2184,
   "ocr_words": 8316,
   "ocr_distinct": 2364
  },
  "recall_percent": 92.6,
  "five_word_phrases_percent": 73.9,
  "control_unrelated_vocabulary_percent": 0.0,
  "critical_values_found": [
   "iso",
   "32000",
   "1993",
   "adobe",
   "postscript",
   "javascript",
   "encryption",
   "acrobat"
  ]
 },
 "resolution_series": {
  "excerpt": "1500 pt tall section of body text",
  "runs": [
   {
    "dpi": 72,
    "words": 860,
    "real_word_percent": 21.3,
    "ocr_seconds": 2.71,
    "image_mb": 0.8
   },
   {
    "dpi": 110,
    "words": 1096,
    "real_word_percent": 92.7,
    "ocr_seconds": 3.13,
    "image_mb": 1.6
   },
   {
    "dpi": 150,
    "words": 1099,
    "real_word_percent": 98.4,
    "ocr_seconds": 4.61,
    "image_mb": 2.9
   },
   {
    "dpi": 220,
    "words": 1099,
    "real_word_percent": 98.7,
    "ocr_seconds": 5.94,
    "image_mb": 5.2
   }
  ]
 },
 "limitations": "Single runs on one document, not averages. Times are orders of magnitude; accuracy figures are stable.",
 "license": "CC BY 4.0 \u2014 reuse with attribution and a link",
 "article": "https://provinglab.dev/measurements/webpage-to-pdf-for-ocr/"
}