Overview of the HASOC Subtrack at FIRE 2021: Hate Speech and Offensive Content Identification in English and Indo-Aryan Languages

• DOI: 10.48550/arxiv.2112.09301
• Publication Date: 2021-01-01
• Authors: Thomas Mandl; Sandip Modha; Gautam Kishore Shahi; Hiren Madhu; Shrey Satapara; Prasenjit Majumder; Johannes Schaefer; Tharindu Ranasinghe; Marcos Zampieri; Durgesh Nandini; Amit Kumar Jaiswal
• Keywords: Offensive; Language identification; Identification

The widespread of offensive content online such as hate speech poses a growing societal problem. AI tools are necessary for supporting the moderation process at platforms. For evaluation these identification tools, continuous experimentation with data sets in different languages necessary. HASOC track (Hate Speech and Offensive Content Identification) is dedicated to develop benchmark this purpose. This paper presents subtrack English, Hindi, Marathi. set was assembled from Twitter. has two sub-tasks. Task A binary classification problem Not Offensive) offered all three languages. B fine-grained classes (HATE) Hate speech, OFFENSIVE PROFANITY English Hindi. Overall, 652 runs were submitted by 65 teams. performance best algorithms task F1 measures 0.91, 0.78 0.83 Marathi, Hindi respectively. overview tasks development well detailed results. systems competition applied variety technologies. performing mainly variants transformer architectures.