Abuse Detection in Online Conversations

This software was designed to detect abusive messages in online conversations. Two main approaches are implemented: a content-based approach and a graph-based approach, which can also be used jointly. This software was applied to a corpus of chat messages written in French, which unfortunately cannot be published due to legal matters. The conversational graphs extracted from this text are publicly available on Zenodo, though.

• URL: https://github.com/CompNet/Alert 
• Production date: 2018–2023
• Related publications:
  • Noé Cécillon, Vincent Labatut, Richard Dufour et Georges Linarès. « Graph embeddings
    for Abusive Language Detection ». In : Springer Nature Computer Science
    2:37 (2021). DOI: 10.1007/s42979-020-00413-7. ⟨hal-03042171⟩
  • Noé Cécillon, Vincent Labatut, Richard Dufour et Georges Linarès. « Abusive
    Language Detection in Online Conversations by Combining Content- and Graph-based
    Features ». In : International Workshop on Modeling and Mining Social Media Driven Complex Networks (Soc2Net). T. 2. Frontiers in Big Data 8. Munich, DE, 2019. DOI: 10.3389/fdata.2019.00008. ⟨hal-02130205⟩ (cite this article if you use the software)
  • Noé Cécillon, Vincent Labatut, Richard Dufour et Georges Linarès. « Tuning
    Graph2vec with Node Labels for Abuse Detection in Online Conversations ». In :
    11ème Conférence sur les modèles et l’analyse de réseaux : approches mathématiques et
    informatiques (MARAMI). Montpellier, FR, 2020. MARAMI ⟨hal-02993571⟩