AlexU-AIC at Arabic Hate Speech 2022: Contrast to Classify
This work addresses cyber-bullying and mental abuse on social media for Arabic-speaking users, but it is incremental as it applies known techniques to a specific dataset.
The paper tackled the problem of detecting hate speech in Arabic tweets by addressing over-fitting in transformer models on small or imbalanced datasets, achieving macro F1 scores of 0.841, 0.817, and 0.476 across three sub-tasks.
Online presence on social media platforms such as Facebook and Twitter has become a daily habit for internet users. Despite the vast amount of services the platforms offer for their users, users suffer from cyber-bullying, which further leads to mental abuse and may escalate to cause physical harm to individuals or targeted groups. In this paper, we present our submission to the Arabic Hate Speech 2022 Shared Task Workshop (OSACT5 2022) using the associated Arabic Twitter dataset. The shared task consists of 3 sub-tasks, sub-task A focuses on detecting whether the tweet is offensive or not. Then, For offensive Tweets, sub-task B focuses on detecting whether the tweet is hate speech or not. Finally, For hate speech Tweets, sub-task C focuses on detecting the fine-grained type of hate speech among six different classes. Transformer models proved their efficiency in classification tasks, but with the problem of over-fitting when fine-tuned on a small or an imbalanced dataset. We overcome this limitation by investigating multiple training paradigms such as Contrastive learning and Multi-task learning along with Classification fine-tuning and an ensemble of our top 5 performers. Our proposed solution achieved 0.841, 0.817, and 0.476 macro F1-average in sub-tasks A, B, and C respectively.