A computer algorithm for Internet troll recognition has been created

Researcher Sergei Monakhov of the Friedrich Schiller University of Jena has developed two computer algorithms for Internet troll recognition. They analyze the tweets of the alleged troll


When creating the algorithm, the researcher used a sociolinguistic approach. The author hypothesized that trolls tend to broadcast a limited amount of information, but publish many posts with different wording. For this reason, trolls have certain repetitive words and pairs of words in the lexicon. The researcher tested his algorithm on the accounts of trolls and real American congressmen. It turned out that the algorithm needed only 50 tweets to determine exactly who was a troll and who was a congressman. The program was even able to distinguish the messages of Donald Trump, which are also provocative, from the messages of trolls.