When the Timeline Meets the Pipeline: A Survey on Automated Cyberbullying Detection

Web 2.0 helped user-generated platforms to spread widely. Unfortunately, it also allowed for cyberbullying spread. Cyberbullying has negative effects that could lead cases of depression and low self-esteem. It become crucial develop tools automated detection. The research on developing these been growing over the last decade, especially with recent advances in machine learning natural language processing. Given large body work this topic, is vital critically review literature within context latest advances. In paper, we survey detection cyberbullying. Our sheds light some challenges limitations field. range from defining cyberbullying, data collection, feature representation model selection, training, evaluation. We provide suggestions improving task addition survey, propose improve by addressing raised limitations: 1) Using contextual models like BERT cyberbullying; 2) slang-based word embeddings generate better representations cyberbullying-related datasets. results show outperforms state-of-the-art deep models. initialized outperform traditional embeddings.