Graph-based Features for Automatic Online Abuse Detection

Etienne Papegnies (1) , Vincent Labatut (1) , Richard Dufour (1) , Georges Linares (1)

1 LIA - Laboratoire Informatique d'Avignon

While online communities have become increasingly important over the years, the moderation of user-generated content is still performed mostly manually. Automating this task is an important step in reducing the financial cost associated with moderation, but the majority of automated approaches strictly based on message content are highly vulnerable to intentional obfuscation. In this paper, we discuss methods for extracting conversational networks based on raw multi-participant chat logs, and we study the contribution of graph features to a classification system that aims to determine if a given message is abusive. The conversational graph-based system yields unexpectedly high performance , with results comparable to those previously obtained with a content-based approach.

Mots clés: Text categorization; Online communities; Abuse detection; moderation

Domaines: Recherche d'information [cs.IR]; Traitement du texte et du document; Réseaux sociaux et d'information [cs.SI]

Etienne Papegnies, Vincent Labatut, Richard Dufour, Georges Linares. Graph-based Features for Automatic Online Abuse Detection. 5th International Conference on Statistical Language and Speech Processing (SLSP), Oct 2017, Le Mans, France. pp.70-81, ⟨10.1007/978-3-319-68456-7_6⟩. ⟨hal-01571639⟩