factual

Unstructured title docs need OCR and NLP

Doma's escrow inputs are often flat image PDFs, so it applies OCR to extract text, then natural language models trained to understand the syntax title professionals use, such as how vesting language implies tasks.

factualpending

Speaker

Max Simkoff

Evidence Quote

“we're training them to understand the syntax of what's being said in the text. And so certain information about the vesting of the property might confer certain tasks”

Source

Max Simkoff - Redesigning the Home Buying Experience - [Founder’s Field Guide, EP. 47]— Invest Like the Best
Created: 11/11/2026, 10:25:15 AM

Your notes

Loading notes...