Practice Bank: Parser Decisions
Use these scenarios to practice production parser judgment.
1 Refund PDF with a broken table
Decide whether the policy can be chunked or needs human review.
2 HTML docs page with navigation noise
Remove boilerplate but preserve headings, anchors, and canonical URL.
3 Scanned incident form
Use OCR confidence and bounding-box evidence to decide readiness.
4 CSV pricing matrix
Preserve column units, customer tier, region, and row identity.
5 DOCX policy with comments
Separate approved policy text from comments and tracked changes.