Forms Extractor Vs. Forms AI

[sharon.palawandram]

Hello Team,

I am developing an automation to handle a structured document with a fixed layout. However, assessing sample data, the structured document comes in various formats, such as native, scanned, skewed, printed and digitized.

Looking at the most feasible extractor to use, Forms Extractor cannot be used due to the variability in layouts (as defined in UiPath documentation).

Can these variabilities be handled in Forms AI?

Forms AI mentions that it can be used for uploading and processing structured forms with standard layouts and fields.

What is the definition of structured forms in Forms AI?

1. Does a structured document’s structure mean the structure of a document? Can it handle native, scanned, skewed scenarios?

2. Or is the structure similar to the structure defined in Forms Extractor?

I am a bit hesitant to move to an ML extractor for this use case, due to time constraints & extraction quantity(50+ fields)

Thank you.

@Monica_Secelean @Anca_Bolca @Vibhor.Shrivastava

[alexcabuz (Alexandru Cabuz)]

Hi @sharon.palawandram

Hard to say without trying it out. However, effort-wise, there is not that big of a difference between Forms AI and ML in this case, since you need to label 10-20 samples either way. I would label 20, see how well Forms AI works. If it does not work well enough, then just export the dataset and train a ML model. With 20 samples, if the layout is the same or very similar (regardless if it is scanned, skewed, etc) it should be enough to get a decent ML model. Then worst case, if it still does not perform, you can add another 10 or 20 samples and it should work very well.

Best,
Alex.

[sharon.palawandram]

Thank you Alex, I will try this methodology.