Hello team,
We are using unstructured IXP model and for a couple of documents the Total amount extracted shows 0 however the reference it picked up seems to be totally fine, the OCR part is not extracting the right value from the reference.
Any idea how to fix this scenario?
You are passing value as 39 410,80 but model except the value 39,410.80 please try to pass the value
Hope it work
Thanks
No, I’m not passing the value manually. It is the Reference field. This is the auto extraction result where reference is fine, but OCR value is 0.
Also the value is 39,410,80 (without any decimal) FYI
In IXP validation (Build tab > Validate predictions), hover over the “Total amount” field showing 0 EUR, click the pencil icon, and manually type the correct value “39.08”. Then click the + icon, re-highlight the exact reference span “39.08” in the document, and submit the annotation.
What is the type of the field (Field Type) and Data Type in the taxonomy?
Try changing the type and predict again for the document.
Please check the type in taxonomy ?
This is expected behavior with unstructured IXP models.
The OCR is reading the value correctly (as seen in the reference), but the model is not confidently mapping it to the Total Amount field, so it returns 0.
This usually happens due to layout differences, multiple amounts on the page, or formatting variations.
It’s a model interpretation issue, not an OCR issue.
Let me try this way to fix it
Hi @arjun.shiroya thanks. I tried this way and trained few docs but still same error. Field type is Monetary Quantity. Also this is only happening for ~5% of docs while for rest of the similar documents its working fine. These all are similarly structured documents.
Try refining your prompt
may be its assuming the value is not as per the prompt
cheers
