Hi guys I have a question.
I want to be able to use generative AI to be able to extract data from the statement document type and I have noticed that the generative extractor doesn’t allow for extraction of line-items, only regular fields. So I’m trying to use the Extract Document Data activity and I’ve set all my prompts.
Firstly is it even possible to use this as an extractor and secondly how are the data type conversions done. The output of Extract Document Data is IDocumentData(Of CustomGPTDocumentType). How do I handle all these conversions if possible since the output of the Data Extraction Scope is ExtractionResult. I want it to be robust so that it caters for normal ExtractionResults and IDocumentData. Thank you!
We recommend you either use the DocumentUnderstanding.Activities or IntelligentOCR Activities - but not together.
To retrieve the Extraction Results from Document Data, you find them available under Properties → ExtractionResult - but again, we don’t recommend the 2 activity packs used together.
Maybe give me some more details about the use case you’re trying to address & I’m trying to look into it/see what is missing from Document Data?
What I’m trying to do is use ‘Extract document data’ activity as a GenAI extractor for statements document type as the ML extractor isn’t performing as well as I want.
The output of Extract Document Data is IDocumentData(Of CustomGPTDocumentType) and the input of present validation station takes in ExtractionResults which is obtained from the output of the Data Extraction Scope. How do I deal with these conversions?
1 Like
I recommend you use just one of the 2 activity packs; you could either
- use the DocumentUnderstanding.Activities with
- a Document Understanding projects consumed from Extract Document Data (same as the one consumed in the ML Extractor currently, if possible)
- Extract Document Data with the Generative Extractor
- Activities - Create Validation Task instead of the Present Validation Action
- use the IntelligentOCR activities with the Generative Extractor instead of Extract Document Data
Hope it helps!
Monica
I have tried the Generative Extractor but it is only able to extract regular fields, it doesn’t allow me to create prompts for the line-items. My goal is to use 1 Extractor for this document type.
We currently don’t support the extraction of tables using our generative fields; however, you could use this as a workaround: https://www.youtube.com/watch?v=BwX5v3UnxfY
Hope it helps,
Monica
