Corpus
What was tested
Updated
Documents listed
0
Downloaded and prepared
0
Pages
0
Needed OCR
0
By category
QuotesProposalsInvoices
| Category | Listed | Downloaded | Pages | Characters | Scanned documents |
|---|
Proposals are few but long: they hold most of the pages. Invoices are short and most often scanned.
How the text was obtained
A page is sent to OCR when its text layer has fewer than about 200 characters or is mostly unreadable characters.
Document length
Most documents are one or two pages. A small number run to dozens of pages and are read page by page.
Duplicates and exclusions
Sampling and coverage counts treat each duplicate or revision pair as one document.