We take DOT plans and turn them into machine-readable objects using:

  1. Region detection trained on plans from 10+ states.
  2. A custom OCR engine fine-tuned on millions of actual DOT plan pages.
  3. Table preprocessing and data caching.
  4. Semantic DOT language understanding using pay-item and work-type language from dozens of states to form a cohesive understanding of the plan using BERT.

Whether it's one plan or 100,000, the process is the same.

Over 30,000 DOT plan sets processed across more than a dozen states — 2.9 million plan pages.

If any of that interests you, reach out for a demonstration:
zac.morton2015@gmail.com
zmmorton@student.ysu.edu

— Jump:
⎘ Copy