SALSA: Detection of Cybertrolls using Sentiment, Aggression, Lexical and Syntactic Analysis of Tweets

Trolls designed to create discord, chaos, and misinformation have been on the rise lately because of the heightened anonymity provided by the internet. Current methods of troll detection focus on identifying whether a user is a troll are based on various features outside a single instance of a post. We explore the possibility of using sentiment, aggression, lexical, and syntactic textual features to determine whether a tweet is meant to troll or not. From experiments and analysis, it was shown that these textual features are indeed sufficient for a classifier to perform well in detecting whether or not a tweet was meant to troll. Furthermore, we also show how these features are relevant for troll detection and how certain ways to combine these features can improve performance in terms of accuracy, precision, recall and F1 score.