Abusive Language Detection in Online Conversations by Combining Content- and Graph-based Features

Noé Cecillon (1) , Vincent Labatut (1) , Richard Dufour (2)

1 LIA - Laboratoire Informatique d'Avignon
2 LS2N - équipe TALN - Traitement Automatique du Langage Naturel

Our objective is to detect abusive messages in online conversations through content- and graph-based methods. Our approach consists in: 1) Exhibiting the importance of structural information in the context of online abuse detection; 2) Combining textual and structural information by proposing fusion methods of both types of information.

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

Noé Cecillon, Vincent Labatut, Richard Dufour. Abusive Language Detection in Online Conversations by Combining Content- and Graph-based Features. Meetup LIAvignon, Nov 2022, Avignon, France. 2022. ⟨hal-03857928⟩