Extacting Invoice Table Data That Spans Multiple Pages
I've been trying to extract invoice line items that sometimes span multiple pages. I'm using 5 documents for the training. Four of these documents contain 3 or fewer line items that fit on one page, and the fifth document has many line items that span 3 pages. While tagging each document, I only tagged the line item data from one page, since tags cannot be used on more that one page of a document. For the large document, I used the second page, since it had the most line items. When testing the resulting model against a document with many line items spanning 3 pages, the resulting JSON only included the line items from the second page.
What can I do differently in order to correctly return all of the line items from all pages of the document?