Skip to content

btisler-DS/ddrp

Repository files navigation

DDRP — Deterministic Document Review Protocol

Version: 0.2.0 Status: Released DOI: 10.5281/zenodo.18427220 Web: quantuminquiry.org/DDRP


About This Repository

This is a reference implementation, not a product.

If you're looking for a turnkey document review solution, this isn't it. DDRP is architectural infrastructure—a working example of how to build deterministic, auditable text analysis without semantic inference. Think of it like publishing a cryptographic library: it doesn't solve your whole problem, but if you need this specific capability (reproducible detection of obligation-creating language with full audit trails), here's a rigorously tested pattern you can adapt.

The Problem

When documents contain obligation-creating language—requirements, definitions, commitments, specifications—how do you verify those obligations were actually addressed?

In high-stakes contexts, this matters:

  • Regulatory submissions that must demonstrate compliance
  • Technical specifications where missing requirements cause failures
  • Legal documents where unresolved obligations create liability
  • AI-generated content where you need to audit what was promised

Current approaches fail at scale:

  • Manual review doesn't scale and isn't consistent
  • Keyword search misses context and relationships
  • AI/ML systems work but can't explain decisions or guarantee reproducibility
  • Semantic analysis is subjective and not auditable

When you need to defend your analysis in court, cite it in policy, or build institutional processes around it, "the AI said so" isn't sufficient.

DDRP's Approach

DDRP detects obligation-creating language using 49 deterministic lexical patterns and tracks whether those obligations get structurally resolved within the document.

Key design principles:

  • No AI/ML — Pure pattern matching, no probabilistic components
  • No interpretation — Detects linguistic structures, doesn't assess meaning
  • Fully deterministic — Identical input always produces identical output
  • Auditable — Every decision traceable to specific patterns and rules

What DDRP detects:

  • REQ operators: "must", "shall", "required to"
  • DEF operators: "defined as", "means", "refers to"
  • CAUSE operators: "because", "due to", "results in"
  • SCOPE operators: "applies to", "limited to", "excludes"
  • UNIV operators: "all", "every", "none"
  • ANCHOR operators: "see section", "defined above", "per attachment"

What DDRP tracks:

  • Whether requirements have corresponding definitions
  • Whether scope is specified for obligations
  • Whether external references are anchored
  • Obligation status: SATISFIED (structurally complete) or OPEN (missing components)

Proof It Works

Determinism guarantee: Identical input + identical ruleset = identical output (byte-for-byte)

Test it yourself:

# Clone and install
git clone https://github.com/btisler-DS/ddrp.git
cd ddrp && npm install && npm run ui:install

# Run hostile audit (26 tests designed to break determinism)
npm run hostile-audit

# Start browser UI
npm run ui

The hostile audit includes:

  • 50 identical runs must produce identical results
  • Semantically equivalent text must produce different outputs (proving it's lexical, not semantic)
  • Pattern changes must predictably affect detection
  • Edge cases designed to expose non-deterministic behavior

If all tests pass, DDRP works as advertised. If not, it's broken.

What This Is and Isn't

This is architectural infrastructure, not a turnkey solution. Think of it like publishing a cryptographic library—it demonstrates that deterministic, auditable document structure extraction is achievable and provides a working reference implementation.

What DDRP does:

  • Detects obligation-creating linguistic patterns
  • Tracks structural completeness
  • Generates reproducible, auditable results
  • Provides cryptographic traceability

What DDRP does NOT do:

  • Interpret meaning or assess correctness
  • Provide legal opinions or compliance determinations
  • Make recommendations or risk assessments
  • Use AI/ML or subjective weighting

Who this is for:

  • Regulatory agencies needing reproducible document analysis
  • Compliance departments that must defend their review processes
  • Research teams requiring auditable text analysis
  • Anyone who needs "because the deterministic pattern matched" instead of "because the AI said so"

Quick Start

Browser UI

npm run ui          # Start at http://localhost:3000

Command Line

npm run verify              # Basic verification
npm run hostile-audit       # Full test suite  
npm run typecheck          # TypeScript validation

Example Output

{
  "protocol": {
    "name": "DDRP",
    "version": "0.2.0",
    "doi": "10.5281/zenodo.18427220"
  },
  "detection": {
    "REQ": [
      {
        "pattern": "must",
        "text": "Users must provide valid credentials",
        "line": 15,
        "position": 42
      }
    ]
  },
  "obligations": [
    {
      "requirement": "Users must provide valid credentials",
      "status": "OPEN",
      "missing": ["definition of valid credentials"]
    }
  ]
}

Project Structure

ddrp/
├── src/                    # Core TypeScript modules  
│   ├── operators_v0_1.ts   # 49 detection patterns
│   ├── obligations_v0_1.ts # Obligation tracking
│   └── ...                 # PDF, hashing, transaction logs
├── tests/                  # Hostile audit (26 tests)
├── ui/                     # Browser interface
└── DDRP_v0_2_SPEC.md      # Full specification

Installation

Prerequisites

  • Node.js v18+
  • npm (included with Node.js)

Setup

git clone https://github.com/btisler-DS/ddrp.git
cd ddrp
npm install
npm run ui:install

Evaluation Questions

Don't ask: "Can DDRP review my documents?"

Ask: "Do I need deterministic, falsifiable document structure extraction as a component in my system?"

If yes:

  1. Run the hostile audit to verify determinism
  2. Examine the 49 patterns to see if they fit your domain
  3. Test with your documents to assess coverage
  4. Fork and modify patterns for your use case

If no, this isn't for you—and that's fine.

Documentation

Document Purpose
DDRP_v0_2_SPEC.md Complete technical specification
ui/public/USER_GUIDE.md Browser UI walkthrough
tests/hostile_audit_v0_2/RUN.md Audit methodology

Citation

@software{ddrp2026,
  title = {DDRP: Deterministic Document Review Protocol},
  version = {0.2.0},
  year = {2026},
  doi = {10.5281/zenodo.18427220},
  url = {https://github.com/btisler-DS/ddrp}
}

License

Released under CC BY-NC 4.0.

Non-commercial use only. See LICENSE for full terms.


Changelog

Version Date Changes
0.1.0 2026-01-29 Initial release
0.2.0 2026-01-30 PDF support, canonicalization, SHA-256, transaction records, DOI

About

Deterministic Document Review Protocol (DDRP) v0.1 — a lexical, deterministic protocol for surfacing structural obligations in documents.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors