How does a language robot decide whether something is racist? This is Textgain, the algorithm that assessed Dries Van Langenhove’s posts | Inland

“Cut your hands, collect diamonds, Congo is ours”, the tone is clear. An algorithm assessed the more than 70,000 messages from chat groups of the extreme right-wing Schild & Vrienden. This shows that the messages from Vlaams Belang MP Dries Van Langenhove were the worst. How does a language robot decide whether a message is racist?