2 ms·
Researchers develop a method to keep bots from using toxic language
- uniqueuid 5y agoI've looked at various networks for detection of incivility/hate speech/insults/profanity, and it's a train wreck. The best models barely reach >.5 true positives in German. And the incivility which they find are the easy, shallow parts. So, even though I have only read the article and not the paper, I predict that this will absolutely, totally not work in any real setting.