Overview of the HASOC Track 2024: Hate-Speech Identification in English and Bengali
Thomas Mandl, Koyel Ghosh, Md Nishat Raihan, Sandip Modha, Shrey Satapara, Tanishka Gaur, Yaashu Dave, Marcos Zampieri, Sylvia Jaki · 2024
Hate speech detection on social media platforms continues to be persistent social issue.Automatic detection is necessary as a first step in content moderation, however, systems are not yet perfect.To continue advancing research on low-resource languages, the shared task Hate Speech and Offensive Content Identification (HASOC) 2024 offered classification tasks in Bengali and in English.A total of 11 teams submitted runs.For English, the leading team achieved an F1 score of 0.813 and for Bengali the highest-performing team achieved an F1 score of 0.716.In HASOC 2024 a large variety of approaches were used by the participants including lexical approaches, transformer-based model as well as zero shot learning with LLMs.