I have a PDF file(single training instance) which can contain many documents(multi-page) within it(invoices, different forms, emails). For e.g a single PDF instance may contain

Page 1 - Overall Cover Page

Page 2 - Form 1(Page 1 of 2)

Page 3 - Form 1(Page 2 of 2)

Page 4 - Form 2(Page 1 of 1)

Page 5 - Cover Page for Form 3

Page 6 - Form 3(Page 1 of 3)

Page 7 - Form 3(Page 2 of 3)

Page 8 - Form 3(Page 3 of 3)

Page 9 - Invoice

Page 10 - Email(Page 1 of 2)

Page 11 - Email(Page 2 of 2)

I have to be able to:

a) classify the documents

b) identify the start and end of each document i.e.

Cover Page - Pg 1 to 1

Form 1 - Pg 2 to 3

Form 2 - Pg 4 to 4

Form 3 - Pg 5 to 8. Note this also includes the cover page at Page 5. However if that's difficult, it can be labelled separately just as Cover Page

Invoice - Pg 9

Email - Pg 10 to 11

Can you point me to literature/open source projects/terminology that I can start reading up on? We have a labelled training set of PDFs.

  • $\begingroup$ Can you clarify what the problem is that you are trying to solve? Are you trying to separate out the forms with a single PDF document? $\endgroup$ – shepan6 Jun 27 at 14:56
  • $\begingroup$ Yes, a single PDF document has many forms/invoices/emails inside it and I want to separate out the forms. Thanks @shepan6 $\endgroup$ – RAbraham Jun 29 at 23:55
  • $\begingroup$ When you say you have a labelled training set of PDFs, do you mean you know where are the boundaries of logical documents in these PDFs? I would suggest starting from one of these approaches: 1) writing rules to determine the boundaries; 2) identify some features of these boundaries, and train a baseline machine learning algorithm (e.g Logistic regression) to see how feasible this is. $\endgroup$ – Albert Jun 30 at 3:44
  • $\begingroup$ @Albert. Thanks. Yes, we have labels for the boundaries of the logical documents within those PDFs e.g. Page 2: Form1 (Page 1 of 2) as in the question. Here, Form1 is the label. In general, it's Page X: Label(Page Y of Z) above. $\endgroup$ – RAbraham Jun 30 at 10:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.