Issue with parsing Large PDFs, and conveting JSON to consistently applied html

[marko30856491]

I'm having issue converting a large PDF, 100+ pages, with images and complex tables (attached PDF and converted html)

Some paragraphs lead with numbers and they are being assigned in separate divs, and are overlapping the paragraphbtext they are assigned to.

There are also large white spaces where headers/footer/page breaks are

Any help appreciated

[marko30856491]

I've attached PDF and output json

[Joel Geraci]

I don't think the structuredData.json belongs with that PDF. The JSON contains a document title "SUMMARY OF PRODUCT CHARACTERISTICS" that does not appear in the document.

[marko30856491]

Hi Joel,

No, it is the correct JSON, the 'Summary of Product Charecteristics' is just a document name and type of document, not necessarily whats contained in the doc itself

Any help appreciated. It is a very complex document, but it is our best test case. We would be converting 2000+ documents per month, and we need 100% accuracy, as it is a regulated space.

Thnaks,

Mark

[Joel Geraci]

Extract doesn't pull metadata from the document. The text "SUMMARY OF PRODUCT CHARACTERISTICS" does not occur on any page in the PDF you uploaded to this thread. That means that the JSON does not belong to that file. I ran your document through Extract and have attached the proper JSON to this response. I've also attached a PDF containing annotations that "visualize" what Extract found. It looks fine to me. How exactly are you interpreting the JSON to create HTML? That might be where the issue is.

[marko30856491]

Hi Joel,

You are right, my problem is in the interpretation of the json. In your example I see that you were able to make page 108 correctly, but in my case it looks like this:

 I'm using php for the html code:

 

Can you guide me a bit about the coordinates and if I should create any other element apart from the "div"?

[marko30856491]

Hi @Joel Geraci have you seen the above? Just wondering if you have any suggestions. Thanks for the help

Mark

[marko30856491]

Hey @Joel Geraci , just checking on the above comments you made. Can you help provide guidance or is there another user in the community who clould if you are too busy? 

Thanks

Mark