AI suggests toned-down alternatives to abusive posts online

Jun 05, 2018, 8:42 AM EDT
(Source: Artotem/flickr)
(Source: Artotem/flickr)

The internet, particularly social media platforms, are notorious for firing a barrage of abusive and filthy slurs from which hapless users have no escape. Researchers at IBM have created an AI algorithm that suggests toned-down alternatives to replace foul language, instead of weeding them out straight away.

The AI system, which learned by sifting through millions of tweets and Reddit posts, shows impressive adeptness at weeding out profane and hateful words, but the suggestions struggle in preserving the meaning of the original comments, reports Mail Online.

Aware of the tool's potential to filter posts critical of governments or online platforms, the researchers want this system to be used only to mellow the tone of coarse messages and not stymie the dissent online, writes Futurism.