Run an extraction system against the supplied PDFs.
OCR AND DOCUMENT AI QA · GROUND TRUTH INCLUDED
Measure extraction accuracy against known synthetic answers.
Test invoices, purchase orders, meeting minutes, and receipts across text-layer and image-only PDFs, then score normalized predictions locally.
$99 one-time purchase
Buy and download nowCommercial license — $249Download free sampleBrowse sample on GitHubCompare all tools
- Automatic delivery
- No subscription
- Private local files
Who it is for
Document-AI developers, OCR evaluators, data-extraction teams, educators, consultants, and technical portfolio builders.
REAL PRODUCT PREVIEW
See what is actually included.
This image was rendered from the delivered product files. Workbook examples and data shown are fictional or synthetic.
EXACT DOWNLOAD SPECIFICATIONS
Know what arrives before checkout.
Archive name, byte size, and file counts are generated from the current packaged download during every site build. Format, requirement, and license labels are checked catalog metadata.
- Personal package
- Document_AI_Extraction_Benchmark_Personal.zip
- Personal size
- 644.01 KiB (659,468 bytes)
- Personal files
- 32 files
- Commercial package
- Document_AI_Extraction_Benchmark_Commercial.zip
- Commercial size
- 644.41 KiB (659,878 bytes)
- Commercial files
- 32 files
- Included formats
- PDF, CSV, JSONL, JSON, Python, Markdown, TXT
- Requirements
- PDF viewer; Python 3 only for the optional evaluator
- License options
- $99 Personal: Personal, educational, portfolio, or sole-proprietor use; $249 Commercial: Commercial use for one organization with up to 10 users
WHAT IS INCLUDED
A focused system for one recurring problem.
- 12 text-layer PDFs and 12 scan-style image PDFs
- Four synthetic business-document classes
- Normalized ground truth in JSONL and flat CSV
- Prediction schema, example output, and local Python evaluator
- Personal and ten-seat commercial-use licenses
DELIVERY
Use it without handing over your data.
The product is delivered automatically after checkout. It is designed for self-serve use and does not include customization, data entry, consulting, or review of your private files.
A SIMPLE WORKFLOW
Start with the next useful record.
Normalize predictions to the included JSON schema.
Score the predictions with the local evaluation script.
Inspect field-level mismatches and repeat the test after changes.
RELATED SELF-SERVE TOOLS
Build the smallest useful system.
Developer and analytics data
Synthetic Business Data Lab
A linked fictional business dataset in CSV, JSONL, and SQLite for Excel, SQL, Power BI, Python, ETL, and dashboard practice.
View product — $29Commercial Windows utilities
Business File Utility Vault
Four read-only Windows utilities for file inventories, snapshot comparisons, filename cleanup plans, and SHA-256 manifest verification.
View product — $49BEFORE YOU BUY
Questions and boundaries.
Are any documents or transactions real?
No. The entire corpus is synthetic and marked as training material.
Does it certify my OCR system?
No. It provides a repeatable extraction benchmark, not legal, financial, accessibility, or compliance certification.
Can a development team use it?
Yes with the commercial license for up to ten organizational users.
BUY ONCE
Download the Document AI Extraction Benchmark.
Sales help fund continued independent transparency work and the software used to preserve and analyze public records.
Get it for $99