Evidence-Grounded Constraint Checking in Construction Documents

2026-07-31

AI Summary

This study addresses the challenge of constraint checking in construction document review, where complex relationships among textual content, geometric elements, page layouts, and cross-version dependencies hinder automation. To tackle this, the authors propose an evidence-based automated review pipeline featuring a rule-aware evidence routing mechanism. This framework integrates a PDF evidence allocator, a four-state deterministic rule engine, Region-RAG retrieval, and overlapping image tiling to normalize factual extraction while preserving source fragments and escalating unresolved cases, thereby enabling traceable auditing. Experiments demonstrate that the optimized image allocation strategy improves accuracy by 10.6 percentage points in project-family standardization decisions and reveal a trade-off between evidence resolution and coverage breadth—challenging the presumed universal advantage of region-focused evidence. However, precise recovery rates remain limited on larger-scale datasets, and false positives persist.

Abstract

Professional-document review is a constraint-checking problem in which decisions depend on relations among text, geometry, pages, and document revisions. We present an evidence-grounded pipeline that normalizes extracted facts, executes four-state rules deterministically, retains source spans, and escalates unresolved cases. We evaluate its PDF evidence allocator on 160 reference-based tasks from 29 construction projects using a repeated four-system test and a disjoint two-system breadth extension. In the repeated test, reallocating a four-image budget from retrieved page overviews to one overview and three overlapping tiles improves project-family standardized decision accuracy by 10.6 percentage points (95% project-cluster bootstrap CI: 4.3 to 18.0; exact p = 0.031). This effect does not persist in the broader block: Region-RAG changes accuracy by -4.1 points (95% CI: -10.2 to 1.9; exact p = 0.209), while an equal-image sensitivity favors page breadth. Exact finding-set recovery remains low, false passes remain common, and repeated-run agreement is poorly calibrated. The results identify a resolution-breadth trade-off rather than a universal advantage for region-focused evidence, motivating rule-aware evidence routing and expert review.

Problem

• constraint checking
• construction documents
• evidence grounding
• document review
• PDF evidence allocation

Innovation

• evidence-grounded reasoning
• constraint checking
• document review automation
• Region-RAG
• rule-aware evidence routing