Single or multiple ML skills for document understanding?

Help AI Center
,
1

which one is better for enterprise purpose with dealing with various kinds of document?

for example, let’s imagine invoices.
each company has own form of invoice.
I made a single ML skill to extract common parts among the invoices.

but as it expands more, it needs other fields too.
or I want to make another ML skill for other common parts.

to do so, I have to label all the fields again.

Single ML skill for common parts from various documents was convenient
at the beginning with faster development.
but it is hard to expand.

So I’m considering making tiny ML skills per invoice type and exporting common parts.
Now maintenance is my concern. there will be too many ML skills.

which one is better and how to balance them?

2

Try using generative extractor with ML skills for other fields that are not extracted by ML

1 Like
3

I considered it too. But it is on-premise so that I can’t use generative extractor.
Also I prefer predefined extractor.