Financial document KV extraction with normalization and reconciliation

[zavera]

Purpose

Adds sample_analyze_financial_documents.py to Pre_or_post_processing_samples.

Covers five IRS form types used in financial aid and tax reconciliation workflows:

• IRS Form 1040
• W-2
• Schedule C (self-employment)
• Schedule E (rental income)
• Schedule K-1 Form 1065 (partnership income)

Post-processing layer normalizes raw Azure DI string output to typed Decimal values and optionally reconciles against reference values with HIGH/MEDIUM/LOW severity scoring. Synthetic sample data included in Data/.

Contributed by Ambreen Zaver, Callisto Tech.

Does this introduce a breaking change?

[x] No

Pull Request Type

[x] Feature

How to Test

pip install azure-ai-documentintelligence python-dotenv
export DOCUMENTINTELLIGENCE_ENDPOINT=
export DOCUMENTINTELLIGENCE_API_KEY=
python sample_analyze_financial_documents.py

What to Check

• Normalization handles currency symbols, parenthetical negatives, percent values, and N/A correctly
• W-2 non-negative field suppression works correctly
• Delta and severity scoring matches expected thresholds

[zavera]

@microsoft-github-policy-service agree company="Callisto Tech"