I'm currently looking at using Document Intelligence to extract data from some forms. These forms, for example, are for requesting assets. As these are smart PDF forms, they have the ability for the end user to dynamically add as many assets as they want via an "Add" button on the PDF form. This means that the asset section may have a set of 1 or more asset related fields for each asset.
For example, this is a form where only one asset is specified:
And this is the form where 2 are specified:
As you can see Asset Name, Number Type and Description are repeated twice, so two asset sections are there in the last example. Again, there could be N number of assets, not just one or two.
Is this situation possible to handle with Document Intelligence? I tried using the Table field type as that's the only option that allows you to specify fields when you don't know them in advance, but the AI doesn't seem to be able to find the fields correctly with a well enough accuracy (after training it with 20+ docs, and I wasn't seeing much improvement after 10 in the accuracy). I'm guessing the AI is expecting an actual table structure, not free-floating fields like in the above example. Is this just not something that can be handled by DI at the moment?
