Skip to main content
Inspiring
Jun 10, 2024 7:00 AM UTC
Solved

How to extract data from multiple pages using Document Extraction AI skill?

I have a field in my form named as description. The description contains value of paragraphs spanning over a couple of pages. While training the model, I selected the data to be extracted from multiple pages by selecting the continue drawing option, but when the model was put into practice, it fetched the data from only one page. Even in the Reconciliation Task, I cannot select data from more than one page. Is there any workaround available for this issue?

sivanagamalleswarit0001
Inspiring
Jun 11, 2024 7:16 PM UTC

Looks like we cannot do selection on multiple pages not even using continue drawing. You can save page 1  as field 1 and page 2 as field 2 and concat them later to display or save accordingly.

Appian recommends using  "Google Cloud Vision Connected System." for paragraph types.

10 replies

stefanhelzle0001
Brainy
Jun 10, 2024 7:53 AM UTC

This approach works best for structured or semi-structured documents. Did you consider to use the new AI extraction features to extract that description in a second step?

umam1720Author
Inspiring
Jun 10, 2024 11:36 AM UTC

I am already using the Document extraction AI skill. I have trained the model as well. The description value that I want to fetch spans over a couple of pages and I want to extract all of that. While training the model I could select the multiple paragraphs in consecutive pages by selecting "continue drawing" on the next page. But in practice, the model only extracts the description from one page, does not fetch from the next page. And, in the reconciliation task, I can only select the description from one page. I want to extract the value of one key- description, which continues on the consecutive pages as well. Will that be possible?

uttrat238161
Participating Frequently
Jun 10, 2024 9:23 AM UTC

Please use Document Extraction AI Skill for this, after training the model with these types of documents you can use this to solve your issue. For reference on how to use it, please check academy.appian.com/

jayaprakashr0001
Participating Frequently
Jun 10, 2024 12:47 PM UTC

Hi [mention:006235f25c724716ae83768f066aea0f:e9ed411860ed4f2ba0265705b8793d05] ,

How many pages you are trying to extract.

umam1720Author
Inspiring
Jun 11, 2024 4:07 AM UTC

Around 2-3 pages

jayaprakashr0001
Participating Frequently
Jun 11, 2024 7:14 AM UTC

After extraction are you using reconsolidation for verifying the data or not. If yes you can  select the correct data to the field.

sivanagamalleswarit0001
Inspiring
Jun 11, 2024 7:16 PM UTC

Looks like we cannot do selection on multiple pages not even using continue drawing. You can save page 1  as field 1 and page 2 as field 2 and concat them later to display or save accordingly.

Appian recommends using  "Google Cloud Vision Connected System." for paragraph types.

umam1720Author
Inspiring
Jun 12, 2024 4:18 AM UTC

That might work, thanks!