PDF is flowable?

[rkomar]

Originally Posted by Elfwreck

FWIW, extracting text *mostly* works. I'd say 85% or more of text-based PDFs (not scans) convert fairly well to Word or HTML formats... and then need cleanup. Remove the headers & page #'s, which extract as just text. Get rid of the forced paragraph breaks at the ends of pages. Find the chapter headers and fix them. (They might be fine. They might be converted to plain text, depending on various font issues.) Look for sets of short lines of text--dialogue especially--that were all crammed into one paragraph.

The text itself tends to extract fine (if there weren't columns or magazine layouts to deal with), but the formatting needs a thorough touchup to be useful.

Right. And all this complexity is only for relatively simple text. Throw in tables, equations,... and the complexity goes through the roof. I remember reading presentations years ago by LaTeX developers who were studying how to apply something like reflow tags to the entire set of elements in a document. None could figure out how to do it reliably, and so it still isn't possible to create reflowable PDF files directly from LaTeX sources. If they can't do it properly from the source files, then it's all but impossible to do it well from the resulting PDF.