On the Impact of Sentiment and Emotion Based Features in Detecting Online Sexual Predators.

According to previous work on pedophile psychology and cyberpedophilia, sentiments and emotions in texts could be a good clue to detect online sexual predation. In this paper, we have suggested a list of high-level features, including sentiment and emotion based ones, for detection of online sexual predation. In particular, since pedophiles are known to be emotionally unstable, we were interested in investigating if emotion-based features could help in their detection. We have used a corpus of predators' chats with pseudo-victims downloaded from www.perverted-justice.com and two negative datasets of different nature: cybersex logs available online and the NPS chat corpus. Naive Bayes classification based on the proposed features achieves accuracies of up to 94% while baseline systems of word and character n-grams can only reach up to 72%.