I have trained a new model and added a label on it that extracts IBAN from an invoice. I edited its properties as “multi-value”, as it is suppose return all IBANs as a list of strings. Nevertheless, it returns only one IBAN eventhough the invoice has many. How to fix it? Should I have more training data that has invoices with more than one IBAN account numbers?
In your labelling dataset in document manager, how often have you labelled documents with more than one IBAN account vs the ones with only one IBAN account?
Can you give some clarity if “every” document has more than one IBAN account location in your document?
One of the reasons could be the model not having sufficient documents labelled with more than one IBAN account.
The training data contains both: invoices with one IBAN and invoices with more than one. Do you suggest that I could do training only with the invoices that has more than one? I wonder if the reason for this problem is that I am using the predefined schema? If it supports extracting only one IBAN…? Eventhough I did change the label value as “Multi value”, as mentioned.