Skip to main content

Posting this so that the powers that be may see it.

Would be great if there were a feature that allows us to force an extracted document to go into validation. 

Why would this be nice?
In my use case I am having to pass dozens of files through the validator every week in hopes to gain higher accuracy in the extraction results. Unfortunately because the DA tool is often confidently grabbing the wrong fields, the extracted document is considered “Success” (even though I know it got the wrong fields) and wouldn’t normally go through validation. I’m having to use alternate methods to force the document into validation each time so that I can train it. I’m currently creating dummy form fields that don’t exist so that the document goes into validation, but with each run I have to come up with a different fake field which is time consuming.

Please educate me if there is a more streamlined way to auto force documents into validation, but this would be a great quality of life feature to have. I know that there is the test mode, but those validation feedback results aren’t actually saved if I remember correctly.


Thank you

Hi ​@Uzumaki,

 

That's weird.

 

Did you try adjusting the confidence score? Particularly by increasing its threshold !!

For sure the fields that you think are having issues would definitely hit when you increase the score.


Hi ​@Uzumaki,

 

That's weird.

 

Did you try adjusting the confidence score? Particularly by increasing its threshold !!

For sure the fields that you think are having issues would definitely hit when you increase the score.

@Padmakumar It doesn’t allow me to adjust or set the confidence threshold for my learning instance. Does this only work on specific document types? I’m using invoices with the ABBYY OCR provider