We make pretty heavy use of Grobid at scite (https://scite.ai) to parse PDFs in order to extract and classify the citation contexts. A great piece of software and I think without it we would be hard pressed to do what we do.
Big shout out to Grobid. It's been backing one project of mine that I've had hosted for a few years now,
It let's you upload and manage academic papers with a super powered text representation that is only possible because of Grobid.
I wrote something similar a while ago https://github.com/agrafix/grabcite — I was surprised how far you could get with just heuristics. The nice thing is that many academic papers are also available with their LaTeX source which avoids a lot of the structure guessing from PDFs.
Great library. I use it to extract the references from scientific papers in my personal project and it performs well even on tricky PDFs.
Could this be used to convert PDFs into audio? I would love to be able to listen to papers while walking / driving
Yes it can, I did just that combined with AWS Polly or some text to speech library of your choosing.
Grobid sounds like "gros bide" in French, meaning either "fat belly" (slang) or "big fail".