AI-driven detection of hate speech on social media: a case study in the French language (HomeCluster Computing)

The understudied field of hate speech identification in French-language social media, where platform ubiquity and user anonymity have increased the frequency of online animosity, is the focus of the research at hand. This work adds to the European linguistic environment by assembling and preparing a bespoke French hate speech dataset through the integration of several corpora, whereas previous research mostly focused on major worldwide languages. After evaluating a variety of binary classification models, such as transformer-based methods, deep learning architectures, and conventional machine learning, DistilCamemBert obtained the highest F1-score of 80%. By employing explainable AI for interpretability analysis, comparative benchmarking, and bias reduction techniques, the study improves the area.

https://link.springer.com/article/10.1007/s10586-025-05553-0