Tips and tricks for creating effective attributes

These tips help you build attributes that return cleaner data and support deeper document analysis.

Creating effective attributes helps Adobe Acrobat Analyzer extract the information you need with accuracy and consistency. Use the following best practices to define, refine, and validate custom attributes for your documents.

Start simple, then refine

Begin with a straightforward description of what you want to extract. Test the attribute before adding conditions or examples.

Example:
“Extract the retainer fee from the document.”

Write clear, plain-language descriptions

Explain the concept as if teaching it to someone new. Clear definitions help Acrobat Analyzer understand what you’re looking for.

Example:
“A rate card lists the roles or positions and the hourly rate for each.”

Use the Evaluate attribute function early and often

Test your attribute on 5–10 documents to see how it performs.

• If the extracted value is too broad, tighten your description or specify the answer format.
• If the attribute expects a number, set the attribute type to Number to enforce it.

Example:
“Return the answer as a percentage only.”

If the description isn't getting the expected result, try adding positive and negative examples

Examples help improve accuracy by showing Acrobat Analyzer what counts as a correct value—and what does not.
Use 3–5 examples to avoid overfitting.

Examples:

• Positive: “A retainer is a percentage paid at the start of the contract.”
• Negative: “Hourly rates should not be treated as retainer fees.”

Provide review feedback

In the evaluation workflow, use the Is this value correct? option to mark results as correct or incorrect. This helps refine the attribute over time.

Cross-check with AI Assistant

If you’re unsure whether your attribute is capturing the correct information, try asking the AI Assistant a direct question. Comparing its answer with your attribute’s extraction can highlight gaps or inconsistencies.

Example:
“For this document, what is the retainer fee?”

Test on a small set before scaling

Start with documents you know contain the information you’re targeting. Once your attribute returns consistent results, expand to a larger set that includes mixed document types.

Troubleshoot "Not Found" results

If there are more "Not found" values than you expect, check some of the following things

• Is the description crisp and clear? Are there any contrary statements?
• Are examples needed?
• Are there any issues with the document itself? (Scanned documents may result in lower quality extractions or missing data. Is the document password-protected and locked? Is the original file corrupted?)
• Is the data you are looking for within the document itself?

Tips:

• Start off broad and give Acrobat Analyzer the benefit of the doubt. Gradually add conditions, criteria, or specific instructions into the attribute description.
• Use AI Assistant and the built-in evaluation workflow to validate the values.
• Explanations are provided within the Eval Workflow on why something was extracted or not. Use this to help refine your attribute description.

For complex attributes, provide step-by-step logic

Break down how the value should be found or interpreted. This is useful for attributes that depend on context or multi-step reasoning.

Example:
Extract the retainer fee by following these steps:

1. Locate the fees section of the contract.
2. Retainer fees appear as percentages at the start of the contract term.
3. Retainer fees at X% are standard; all other percentages indicate negotiated rates.
4. Return the percentage and whether it is standard or negotiated.
5. Answer format: “Retainer fee %; standard or negotiated.”