Research paper 10

From Public Records to Publication Claims: A Human-Reviewed Evidence Promotion Pipeline

Public-record analysis is often presented as a polished narrative with citations. That format hides the transformation from source artifacts to extracted units, machine-organized candidates, human-reviewed findings, quote checks, procedural gates, and blocked stronger claims. This paper describes a public-evidence promotion pipeline in the WLG East Crazy corpus. The validated source substrate contains 1,358 source artifacts, 24,756 source units, 4,255 PDF page render fingerprints, 45 OCR text units, and zero source-unit validation failures. From that substrate, human review promotes 275 publication findings and records 92 accepted nonpromotion decisions. A reader-facing posture compresses the promoted material into seven issue lanes, seven selected quote trails, and eight blocked stronger claims. A PAE gate layer records 12 decision points with outcomes BLOCK 2, ESCALATE 6, and PERMIT 4. The contribution is a claim-promotion method for public-record intelligence: publication findings must remain traceable to source units, review decisions, quote checks, and external-use gates.