#pdf extraction

Reading 28 Chinese E-Invoices Without OCR: 83% Becomes 100% When You Check the Uppercase Total

On 28 invoice PDFs: numbers parsed 28/28, amount accuracy ranged 25% to 100% by method, and 39.3% of the files were duplicates.