N-grams, wordbags. Filtering inflammatory phrases/ words commonly found in disinformation or toxic IM/social media posts. Impossible to use people to search through entirety of social media posts - too many posts, necessary to use algorithms to filter out the most obscene messages. Must be careful to monitor quirks in the algorithm so it doesn't arbitrarily restrict new phrases or terms which enter our language without considering context. Human moderation still needed to adjudicate on disputed matters.