CCA-F: Claude Certified Architect — Foundations

By Anthropic · Question Mode

QuestionQ104

Context Management & Reliability

You are developing a structured data-extraction system with Claude. The system pulls information from unstructured documents, validates its output with JavaScript Object Notation (JSON) schemas, and preserves high accuracy. It must gracefully handle edge cases and integrate with downstream systems.

Your extraction pipeline processes invoices and extracts line items, subtotals, tax amounts, and grand totals. During evaluation, you find that in 18% of extractions, the total of extracted line-item amounts does not match the extracted grand total—sometimes because of OCR errors in the source document and sometimes because of model extraction errors. Downstream accounting systems reject records whose totals do not match.

What is the most effective way to improve extraction reliability?

Explanation

A deterministic calculated total derived from the extracted line items can be compared with the separately extracted stated total. Retaining both values and routing discrepancies to human review detects both OCR and extraction errors while preserving the source evidence and avoiding fabricated financial adjustments before records reach the accounting system.

Community Discussion

No comments yet. Be the first to start the discussion!