For the subset of PDFs that are invoices, there's the data extractor project called invoice2data, which implements many of the ideas presented in the article.

https://github.com/invoice-x/invoice2data