Document pipelines

Extraction of fragments

→

Fragment order

Classification of fragment

Table parsing

→

                        family name,name,company
                        Rasmussen,Mads,QAECY
                        Dohmen,Philipp,QAACY
                        Argyris,Manos,QAECY
                    

Machine can do maybe 90 %

Post proceessing: Validation/correction

Features of interest

Company, Person, Address, ...

Other special entities?

Is A. Einstein the same as Albert Einstein?

Post proceessing: Collation

Document summaries by vision models