According to the researchers, Twitter trolls are immediately noticed and captured in just 50 posts with the help of an algorithm. This is achieved by "distinguishing" word repetitions made up of repeating words
Internet trolls are among the most significant problems of social media. They can attack others for no reason, create unnecessary discussions, and often deliberately spread false/erroneous content.
According to scientists, it is possible to find and extract Twitter trolls only by reviewing 50 Tweets. The basis of this study is the use of "distinctive" repetitive words used by trolls.
The algorithm that discovers language patterns
The algorithm first defines language patterns and usage. Then he separates the prominent trends in troll messages from the language structure in other words. Thus, he learns the difference between troll messages and daily messages.
According to the study, trolls start using the same words or word patterns repeatedly over time. The reason for this is shown as trying to affect more than one account at the same time.
In previous studies, studies were carried out by considering the trolls' shipping times, the labels they use, and their geographical location. On the other hand, the number of studies taking into account the relationship between language use and Tweets was very few.
The spread of the same patterns betrays the trolls.
According to the researchers, two different variables are generally effective in catching trolls. The first is the repetitive word usage, and the other is the repeated use of similar words by more than one account.
Thanks to the mathematical model developed by considering the repetitive word ratios used and the general usage, it is revealed in only 50 posts whether the accounts are trolls or not.
Trolls, whose purpose is to influence or distort people's perceptions, were on the list after the USA's research in 2018. It was alleged that Russian trolls influenced the US elections.