Smart Metadata for Purchase Order Number

Smart Metadata learns to extract values based on the context in which they appear in the documents, so the more similar the different incoming documents are the higher the probability that the correct value can be extracted. You would need to test this with real documents to find out for sure if it works in all cases.

However, if the PO numbers always follow a certain pattern, for instance if they always have "PO" in the beginning, then you might want to use Text Analytics instead since it can detect such patterns in the content more reliably.

Smart Metadata learns to extract values based on the context in which they appear in the documents, so the more similar the different incoming documents are the higher the probability that the correct value can be extracted. You would need to test this with real documents to find out for sure if it works in all cases.

However, if the PO numbers always follow a certain pattern, for instance if they always have "PO" in the beginning, then you might want to use Text Analytics instead since it can detect such patterns in the content more reliably.

Hi Joonas,

Meaning to say, there is some possiblities that the PO number can still be extracted despite of the different format from different client?

I think that's possible at least in theory if the documents would usually contain similar data near the PO number, for instance an order number and an order date or something like that, so that the context can be identified. But as said, you would need to test this with real documents and different PO number formats.

Hi Joonas, 

Following the question above, i would like to ask whether M-Files Smart Metadata able to train documents in Excel format? I have tried uploading more than 100 excel documents, however it showed error indicating not enough sample. If im using words documents, it worked fine

My understanding is that in general Smart Metadata should work with any document that has machine-readable text content, but the actual behavior will of course depend on the specific documents you are working with: their similarity and so on. M-Files Smart Metadata - FAQ has some further information on why the training might sometimes not be successful even with enough sample documents.