You might consider converting the pdf files to rtf, there are command line tools that do that. Once it is converted to an rtf, you can programmatically read and parse through the text. I am assuming that your pdf files contain tables, but even if they don't this will still probably work.
How clear are the PDFs, have you tried any OCR?
Thank you guys! My pdf are very UNCLEAR! they mix with photo they are like 30 MO each minmum a nightmare...
Plus it's not realy strutured like, every paragraph start with a name of a product. The document is really made to be read by a seller and not to a computer.
I will google OCR and RTF to chose witch one is the best.
Do you guys(and girls) have idea how big company like amazon deal with that?
Mechanical Turk
Are they single page PDFs? If so, you might want to first convert them to high quality JPEGs, since that will be easier to deal with. With Mechanical Turk you'll want to make a task that gives the worker one image at a time. Instead of working with Excel sheets, have you consider using Google Forms? You can give access to a Form and not to the underlying Sheet itself. If you use a Google Sheet as your "system of record" you will automatically have it backed up and also you can always have the option to export to Excel if you need to.
I think MS Office 2016 can convert PDF to RTF also.