Abstract: Extracting key-fields from a variety of document types remains a challenging machine learning problem. Services such as AWS and Google Cloud provide text extraction products to ""digitize"" images or pdfs. These return phrases, words and characters with their corresponding coordinate locations. Working with these outputs remains challenging and unscalable as different document types require different heuristics. The speed-limit for extracting information remains in the ~3-5+ seconds range, too slow for utilizing video and AR to present results
We propose a compressed on-device solution that extracts fields in sub-second range, with speeds approaching 200ms on modern devices for field extraction from invoices and receipts. Real-time scanning and extraction opens up new possible product workflows.
Bill.com is working to build a paperless future. We parse through millions of documents a year ranging from invoices, contracts, receipts and a variety of other types. Understanding those documents is critical to building intelligent products for our users.
Bio: Eitan is the Chief Data Scientist at Bill.com and has many years of experience as a researcher. His recent focus is on machine learning, deep learning, applied statistics and software engineering. Before, he was a Postdoctoral Scholar at Lawrence Berkeley National Lab, received his PhD in Physics from Boston University and B.S. in Astrophysics from University of California Santa Cruz. Eitan holds 4 patents and 11 publications to date and has spoken about data at various conferences around the world.