JusText seems to remove too much and its accuracy seems not to be very high. We need a better tool for boilerplate removal.
Options:
The first two options (and the last one as well, probably) use DL, and need training data. For now, let's experiment with Trafilatura.