Currently, I am learning about OCR. I have a problem about invoices that require OCR to extract tables from a PDF file. But when I use PaddleOCR, there are many text parts that are wrong or missing and the table structure is not like the original file. Please give me some ideas. Thanks