Your team is extracting structured data from 50,000 legacy legal contracts under a two-week deadline.…

Question

Your team is extracting structured data from 50,000 legacy legal contracts under a two-week deadline. Initial testing with 500 sample documents shows 82% pass JSON schema validation on the first attempt, while the remaining 18% fail due to diverse issues—missing required fields, malformed dates, and incorrectly identified parties. Documents that fail typically need refinements targeting their specific failure modes before extraction succeeds. Which batch processing strategy is the most cost-efficient while still meeting the deadline?

Accepted Answer

B. Submit all 50,000 documents via batch API, then submit failed extractions in successive batches—refining prompts between each batch—until all documents pass validation.

Answer

A. Split documents into 10 sequential batches of 5,000 each, analysing results and refining prompts between batches to improve extraction quality progressively.

Answer

C. Use the real-time API for all 50,000 documents since the batch API's 24-hour processing window creates unacceptable deadline risk.

Answer

D. Process 2,000 sample documents via real-time API to identify failure patterns and refine prompts, then batch process all 50,000 with the optimized prompts.

Question 220