I think what we most need is some clarity over certain if this then that conditions where we can work to train a proper classifier.
To this point, our different places have been pretty all-over-the-place and so the work we do is just too scattered for us to have high confidence in exactly what to tweak.
I feel like the most effective solution might be to do spam detection through a similar pattern to how we do human work— marking something as spam and then letting someone confirm or invalidate the selection. That way at least there is clarity and parity in the process vs the black box.
For further actions, you may consider blocking this person and/or reporting abuse
I think what we most need is some clarity over certain
if this then thatconditions where we can work to train a proper classifier.To this point, our different places have been pretty all-over-the-place and so the work we do is just too scattered for us to have high confidence in exactly what to tweak.
I feel like the most effective solution might be to do spam detection through a similar pattern to how we do human work— marking something as spam and then letting someone confirm or invalidate the selection. That way at least there is clarity and parity in the process vs the black box.