[cafard]

I was very impressed with "Camelot" (https://camelot-py.readthedocs.io/en/master/). My impression was that it extracted maybe 80 or 90% of the text properly, far better than anything else I had tried.