neochart v2 research

xref/out/r2/audit_v2_summary.json

Source: xref/out/r2/audit_v2_summary.json

Exact source text

{
  "pages": 2000,
  "population": 2023221,
  "phase_one_frame": 10000,
  "strata": 915,
  "counts": {
    "act_anaphor:resolved": 32,
    "act_anaphor:unresolved": 1173,
    "case_full:resolved": 39166,
    "case_full:unresolved": 3,
    "case_id:resolved": 2222,
    "case_id:unresolved": 3040,
    "case_short:resolved": 5788,
    "case_short:unresolved": 2845,
    "cfr:resolved": 534,
    "named_act:resolved": 89,
    "named_act:unresolved": 3051,
    "session_law:resolved": 335,
    "statute_full:resolved": 2577,
    "statute_short:resolved": 328,
    "statute_short:unresolved": 11608
  },
  "new_patterns": [
    "case_short",
    "case_id",
    "session_law",
    "cfr",
    "named_act",
    "act_anaphor",
    "statute_short"
  ],
  "hand_review": "completed; precision_review.json and recall_gold.json (external QA1 pending)",
  "recall": {
    "detection_recall": {
      "estimate": 0.8438320869867246,
      "page_bootstrap_95": [
        0.7074018923921565,
        0.9297112459916663
      ],
      "valid_bootstrap_draws": 2000,
      "weighted_successes": 2252.5707384403036,
      "weighted_events": 2669.453761214631
    },
    "resolution_recall": {
      "estimate": 0.19055686933187962,
      "page_bootstrap_95": [
        0.040265838578918675,
        0.5882467708410914
      ],
      "valid_bootstrap_draws": 2000,
      "weighted_successes": 431.57073844030367,
      "weighted_events": 2264.787094547964
    }
  },
  "resolver_sha256": "ee25f313096dbd2dd9102047711f9a50328e81d81c1cc954b80c4469b20cc686",
  "ci": "review_metrics_r2.py uses weighted page-cluster bootstrap after manual labels, conditional on phase-one frame",
  "design": "Two-phase sample: SRS manifest frame, then actual court/decision-decade stratification. Estimates are conditional on phase-one frame and omit phase-one variance; this limitation is explicit."
}