DART (Digital Accessibility Remediation Tool)

A toolkit for converting PDF documents to WCAG 2.2 AA compliant accessible HTML.

Features

• pdftotext extraction for born-digital PDFs (best quality)
• OCR fallback for scanned documents using Tesseract
• Automatic structure detection - headings, paragraphs, tables, references
• Image extraction - raster images and vector graphics rendered as PNG
• AI-powered alt text - automatic image descriptions using Claude
• WCAG 2.2 AA validation - built-in accessibility validation
• WCAG 2.2 AA compliant output including:
  • Skip links for keyboard navigation
  • ARIA landmarks and roles
  • Semantic HTML structure
  • Dark mode support (via prefers-color-scheme)
  • Reduced motion support (via prefers-reduced-motion)
  • Focus not obscured (WCAG 2.2 - 2.4.11, 2.4.12)
  • Focus appearance compliance (WCAG 2.2 - 2.4.13)
  • Target size minimums (WCAG 2.2 - 2.5.8)
  • Responsive design
  • Print styles

Installation

System Dependencies

# Ubuntu/Debian
sudo apt install poppler-utils tesseract-ocr

# macOS
brew install poppler tesseract

Python Package

cd DART

# Basic installation
pip install -e .

# With OCR support
pip install -e ".[ocr]"

# Full installation (all features)
pip install -e ".[full]"

Usage

Command Line

# Basic conversion
python convert.py document.pdf

# Specify output directory
python convert.py document.pdf -o ./accessible/

# With verbose output
python convert.py document.pdf -v

# For scanned PDFs with OCR
python convert.py scanned.pdf --dpi 400

Python API

from pdf_converter import convert, PDFToAccessibleHTML

# Simple conversion
result = convert('document.pdf')
if result.success:
    print(f"Saved to: {result.html_path}")
    print(f"Title: {result.title}")
    print(f"Pages: {result.pages_processed}")

# Custom configuration
converter = PDFToAccessibleHTML(dpi=400, lang='deu')
result = converter.convert('german_doc.pdf', output_dir='./output/')

WCAG Enhancement Only

If you already have HTML and want to enhance it for accessibility:

from pdf_converter import enhance_html_wcag, WCAGOptions

html_content = open('document.html').read()
enhanced = enhance_html_wcag(html_content, WCAGOptions(
    add_skip_link=True,
    dark_mode=True,
    add_aria_landmarks=True,
    wcag_version="2.2"  # Target WCAG 2.2 AA
))

WCAG Validation

Validate existing HTML against WCAG 2.2 AA:

from pdf_converter import validate_html_wcag, WCAGValidator

# Simple validation
report = validate_html_wcag(html_content)
print(f"Compliant: {report.wcag_aa_compliant}")
print(f"Issues: {report.total_issues}")
print(report.to_text())

# Detailed validation with file
validator = WCAGValidator(strict_mode=True)
report = validator.validate_file('document.html')
print(report.to_json())

Examples

See the examples/ directory for conversion examples:

• NASA NESC 2024 (examples/nasa_nesc_2024/) - Flagship example with a 76-page multi-column NASA publication. Includes detailed workflow documentation.
• Small Samples (examples/small_samples/) - Quick reference examples including math textbooks, academic papers, and image-only documents.

Gold Standard Template

The templates/gold_standard.html file demonstrates the target output format with:

• Complete semantic HTML5 structure
• ARIA roles and landmarks
• Accessible tables with proper headers
• Skip links for keyboard navigation
• Dark mode CSS support
• Responsive design
• Print-friendly styles

Use this as a reference when customizing output styles.

Directory Structure

DART/
├── convert.py              # CLI entry point
├── pyproject.toml          # Package configuration
├── README.md               # This file
├── CLAUDE.md               # Development guidance
├── examples/               # Conversion examples and workflows
├── pdf_converter/          # Core package
│   ├── __init__.py         # Public API
│   ├── converter.py        # PDF to HTML converter
│   ├── wcag_enhancer.py    # WCAG 2.2 enhancement
│   ├── wcag_validator.py   # WCAG 2.2 validation
│   ├── cli.py              # CLI implementation
│   ├── alt_text_generator.py  # AI-powered alt text
│   ├── image_extractor.py  # PDF image extraction (raster + vector)
│   ├── embed_images.py     # Image embedding utility
│   ├── math_processor.py   # Math/LaTeX processing
│   └── claude_processor.py # Claude API integration
├── templates/
│   └── gold_standard.html  # Reference template
├── output/                 # Default output directory
└── tests/
    ├── test_alt_text_generator.py
    ├── test_image_extractor.py
    └── test_math_processor.py

CLI Options

Usage: python convert.py <input.pdf> [options]

Options:
  -o, --output DIR       Output directory (default: ./output/)
  -n, --name NAME        Output filename (default: derived from PDF)
  --no-dark-mode         Disable dark mode CSS support
  --no-ocr               Skip OCR fallback for image-heavy PDFs
  --dpi N                DPI for OCR processing (default: 300)
  --lang CODE            Tesseract language code (default: eng)
  --no-vector-graphics   Skip vector diagram detection/rendering
  --vector-dpi N         DPI for vector graphics rendering (default: 150)
  --vector-min-drawings  Min drawing ops to detect vector region (default: 5)
  -v, --verbose          Enable verbose output
  --version              Show version information

WCAG 2.2 AA Compliance

The generated HTML meets WCAG 2.2 AA success criteria:

Core WCAG 2.1 Criteria

Criterion	Description	Implementation
1.3.1	Info and Relationships	Semantic HTML, ARIA landmarks
1.4.1	Use of Color	Links underlined, not color-only
1.4.3	Contrast (Minimum)	4.5:1 text contrast ratio
1.4.10	Reflow	Responsive design, no horizontal scroll
2.1.1	Keyboard	All interactive elements keyboard-accessible
2.4.1	Bypass Blocks	Skip link to main content
2.4.2	Page Titled	Meaningful page title
2.4.6	Headings and Labels	Descriptive headings
2.4.7	Focus Visible	Clear focus indicators
2.3.3	Animation	Reduced motion support

New WCAG 2.2 Criteria

Criterion	Description	Implementation
2.4.11	Focus Not Obscured (Minimum)	scroll-margin for fixed elements
2.4.12	Focus Not Obscured (Enhanced)	scroll-margin-top/bottom: 80px
2.4.13	Focus Appearance	3px outline (exceeds 2px minimum), 3:1 contrast
2.5.8	Target Size (Minimum)	24x24px minimum for interactive elements

License

MIT License