Skip to content

Latest commit

 

History

History
21 lines (12 loc) · 1.1 KB

confidence.md

File metadata and controls

21 lines (12 loc) · 1.1 KB
description
What you need to know about confidence

How to deal with uncertain predictions

End-to-end confidence

Every field the model extracts has a corresponding confidence value. The confidence is different from a traditional OCR confidence in that it does not only estimates the probability that the characters are interpreted correctly, but also that it has extracted the correct information (e.g. the total amount and not the VAT amount).

The figure shows example predictions together with confidence values

End-to-end confidence increases automation

You can trust that the model is correct when it says so.

When the confidence of a prediction is above a given threshold, the field can be hidden from the human validator.

This ensures that that only fields that the AI is uncertain about will be manually inspected, while the rest of the fields are fully automated. This means that users will save time and cost by not having to validate high-confidence predictions!

The figure shows how confidence can be used to automate validation of data extraction