Leveraging Affective Bidirectional Transformers for Offensive Language Detection
This work addresses the need for safe online experiences by improving detection of offensive and hate speech, but it is incremental as it builds on existing affective models and shared task benchmarks.
The paper tackled the problem of detecting offensive and hate speech on social media by developing deep learning systems that fine-tune affective models for data augmentation, achieving 89.60% accuracy for hate speech and 95.20% accuracy for offensive language on test data.
Social media are pervasive in our life, making it necessary to ensure safe online experiences by detecting and removing offensive and hate speech. In this work, we report our submission to the Offensive Language and hate-speech Detection shared task organized with the 4th Workshop on Open-Source Arabic Corpora and Processing Tools Arabic (OSACT4). We focus on developing purely deep learning systems, without a need for feature engineering. For that purpose, we develop an effective method for automatic data augmentation and show the utility of training both offensive and hate speech models off (i.e., by fine-tuning) previously trained affective models (i.e., sentiment and emotion). Our best models are significantly better than a vanilla BERT model, with 89.60% acc (82.31% macro F1) for hate speech and 95.20% acc (70.51% macro F1) on official TEST data.