Form recognizer and non-us invoices without pre-defined structure: best practice?

Question

Hello everybody!

I would like to use the form recognizer service to analyze invoices. However, the invoices don't follow a predefined structure.
Do you have any similar cases? are there any best practices you know to handle such situations? Would composite models work in this case? i.e. does the cognitive service "learn" a bit from each and tries to find a solution accordingly or it finds which of the models forming the "composition" best suits the case and then uses it?

Thank you and best regards,

Vittorio

Answer 1

Hello,

I think you are asking custom model for your business. Besides prebuilt model, you can use custom model. A custom model is a machine learning program trained to recognize form fields within your distinct content and extract key-value pairs and table data. You only need five examples of the same form type to get started and your custom model can be trained with or without labeled datasets. Custom models are trained for your distinct data and use cases.

Reference: https://learn.microsoft.com/en-us/azure/cognitive-services/form-recognizer/label-tool
https://learn.microsoft.com/en-us/azure/cognitive-services/form-recognizer/concept-custom?tabs=fott

Regards,
Yutong