Hi @SKT
Welcome to Microsoft Q&A! Thanks for posting the question.
Based on the response from your action of Form Recognizer you can use create conditional statement that control workflow action in logic app.
Update 7/5:
The Analyze Invoice action returns InvoiceAnalyzeOperationResult which has multiple properties and analyzeResult property is one of the response with this output.
Behind the scene logic app uses the form recognizer API and returns the response. I have tested with the sample input and I don't see any field that can mapped directly for your scenario. There is no functionality to confirm confidence score from the service perspective. For form recognizer perspective it is basically a form where you are training a custom model for a particular form type. It will simply extract fields and provide an output in JSON. The analyze invoice is a prebuilt model to extract certain fields it has been trained on.
Basically if you are receiving fake invoices the processed data from the API should be validated in your application. It is easy to fake a format and send a similar form but your apps needs to validate this.
From logic app perspective there are no connectors that can help in validating fake invoices.
Please feel free to raise a feature request on UserVoice to support it, if this is something that you would want to see in form recognizer.
Feel free to get back to me if you have any queries or concerns.
Please 'Accept as answer' and ‘Upvote’ if it helped so that it can help others in the community looking for help on similar topics.
@SKT The Analyze Invoice action returns InvoiceAnalyzeOperationResult which has multiple properties and analyzeResult property is one of the response with this output.
Behind the scene logic app uses the form recognizer API and returns the response. I have tested with the sample input and I don't see any field that can mapped directly for your scenario. But I will check with the form recognizer experts to comment if this can be achieved.
@SKT I have got the confirmation from form recognizer team that there is no functionality to confirm confidence score from the service perspective. For form recognizer perspective it is basically a form where you are training a custom model for a particular form type. It will simply extract fields and provide an output in JSON. The analyze invoice is a prebuilt model to extract certain fields it has been trained on.
Basically if you are receiving fake invoices the processed data from the API should be validated in your application. It is easy to fake a format and send a similar form but your apps needs to validate this.
From logic app perspective there are no connectors that can help in validating fake invoices.
Please feel free to raise a feature request on UserVoice to support it, if this is something that you would want to see in form recognizer.
Let me know if you have any queries or concerns.
Thanks Mayank. That answers my question.
@SKT Thanks for your response and providing more clarification on your comment. I have updated my answer with our discussion. Feel free to 'Accept as answer' so that it can help others in the community looking for help on similar topics.
Hi Mayank,
Fake invoices are another problem which is handled by typical business processes such as purchase order numbers.
I'm just trying to avoid trying to process any other random attachments that comes through such as typical spam that seems to end up in the accounts dept mailbox.
It would be great if there was a analyzeResult property that said something like:
If there is a Purchase order, and the text "invoice" exists, and there is a "due date", and an "amount payable" = isInvoice = 80% certain
(or something like that)
I guess I can write it myself, but would be nice if it was there built into the analyze invoice.
Thanks for your digging into this for the community though.
@SKT Thanks for your feedback and I will pass your feedback to the form recognizer team. You can also create the feature request so it can be tracked, followed up and others can also vote them if they are looking for the same feature. Product team are continuously monitoring these feedback
Sign in to comment