Improve extraction of contents inside tables on documents

Improve reading table structures on invoices or similar documents and correctly extracting and mapping their contents. Currently, when the table structure is complex, doesn't have clear headers and/or contains a lot of text, the extraction quality isn't sufficient.