vamseeachanta Yield Reality (WRK-1246 Corpus Assessment)
Yield Reality (WRK-1246 Corpus Assessment)
WRK-1246 assessed deep extraction yield across 420K+ text-extractable documents. Only two content types have proven extraction yield from the current parsers:
| Content type | Yield | Status |
|---|---|---|
| tables | 69-93% | Production-ready — primary extraction target |
| figure_refs | 1-52% | Partial — metadata only, varies by stratum |
| equations | 0% | Not yet implemented — parsers do not reliably detect |
| constants | 0% | Not yet implemented — parsers do not reliably detect |
| procedures | 0% | Not yet implemented — parsers do not reliably detect |
| worked_examples | 0% | Not yet implemented — parsers do not reliably detect |
These content types are NOT currently extractable from the corpus — they exist in the manifest schema (documented below) but the parsers do not reliably detect them. The schema is retained for future parser development.