The answer highly depends on the nature of your PDFs and the result you want!

There is no simple answer for this general question, because a pure print format and a flowing format are different by nature and (as already mentioned) need heuristics.

This post lists some possibilities (especially pdftohtml -xml) and other corresponding discussions:

Parsing PDFs by text position?