This study investigates the challenges in using linguistic approaches for internet troll detection by analyzing over 180,000 app reviews. The research reveals that a "troll coefficient," previously effective in identifying trolls, also flags genuine negative reviews. This phenomenon is attributed to collaborative repatterning in online discussions where users adjust their language to avoid repetition, mimicking troll behavior. The study proposes a model to predict troll coefficients based on time intervals between reviews, suggesting that the previously successful algorithm needs refinement to account for this collaborative effect.
Publisher
HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS
Published On
Nov 18, 2021
Authors
Sergei Monakhov
Tags
internet trolls
troll coefficient
linguistic analysis
app reviews
collaborative repatterning
troll detection
negative reviews
Related Publications
Explore these studies to deepen your understanding of the subject.