[h_r]

Seems like that would require an understanding of language that far exceeds the current state of the art. And much of the time logic doesn't even apply because the text is incoherent.

[pydry]

You don't need that. You can use humans to flag and verify logical fallacies. The computer can just be tasked with labeling.