IPP accuracy benchmark: sample documents (set v1, seed 42)
==============================================================

These 17 files are taken unchanged from the synthetic test set behind
https://invoiceparserpro.com/accuracy. Every vendor, customer, number and
amount in them is invented. Town names are real places. No customer
document was read, copied or imitated to make them.

expected.json holds the right answer for each file, so you can run the
files through any invoice reader (IPP or another) and score it yourself.

How the published benchmark scores a reading
--------------------------------------------
- Vendor name: equal after dropping case, punctuation and legal-form
  words (Pty, Ltd, LLC, GmbH...), or close enough (similarity 0.85 or
  more), or one name contains the other.
- Invoice number: equal after removing case, spaces, '#' and a leading
  label. Where none is printed, right only when left empty (a PO number
  in its place is wrong).
- Dates: the ISO date. The due date is scored only where one is printed.
- Currency: the ISO code. On a dual-currency invoice, the currency of the
  "amount payable" box.
- Subtotal, tax, total: exact to the cent. The subtotal is the net
  amount, before VAT, on VAT-inclusive layouts too. Credit notes compare
  magnitudes.
- Document type: invoice, credit_note, or other (a supplier statement is
  right only as other).
- Line items: each expected line matched by amount (within a cent) and
  description; a document's lines are right when every line matches.
- should_flag: true for an invoice whose printed figures do not add up,
  and for a file holding several invoices. A reader should flag those for
  a person instead of correcting or accepting them.

A document is "fully correct" when every scored key header field is
right: vendor, number, dates, currency, subtotal, tax, total and
document type (scored_fields lists what applies to each file).

Several invoices in one file are scored against the first invoice.

Files: clean-de-grid-eu-decimals.pdf, clean-za-band-space-groups.pdf, currency-dual-usd-zar-1.pdf, currency-nad-bare-dollar.pdf, tax-inclusive-total-na.pdf, tax-sage-subtotal-after-vat-1.pdf, credit-brackets-gb.pdf, multipage-3p-na-3.pdf, freight-clearing-na-1.pdf, freight-carrier-za-3.pdf, hard-total-inflated.pdf, hard-multiple-invoices-2.pdf, hard-statement-za.pdf, handwritten-invoice-book-poor-6.jpg, photo-motion-blur-severe.jpg, photo-rotate-small-severe.jpg, scan-fax-severe.png
