Issue with Multi-Line Field Extraction in Appian IDP – Training Shows 100% Recall, but Testing Fails

Hi everyone,

I'm encountering an issue with Document Extraction in Appian Intelligent Document Processing (IDP). Specifically, I'm working on extracting fields that can contain multiple lines of text.

Here’s what’s happening:

• I trained my extraction skill on several documents.

• The training report shows 100% recall and precision for the field in question.

• However, when I test the skill with the same or similar documents, the multi-line field is not extracted at all (like descriptions).

• This only happens with multi-line fields — other fields (like dates, names) extract fine.

 What I’ve tried:

• Adjusted the label zone and verified it's correctly mapped to the field.

• Ensured the training data included examples with line breaks and variable-length text.

• Tried with both scanned PDFs and text-based PDFs.

• Reviewed field settings in the AI Skill configuration.

 My question:

Has anyone faced a similar issue where a multi-line field trains perfectly but fails during testing? Is there a trick or workaround for extracting free-text fields with line breaks in Appian?

Would love any insights or experiences — thanks in advance!