Neural models for abusive language detection
User-generated text on social media often suffers from a lot of undesired characteristics, including hate speech, abusive language, insults, etc. that are targeted to attack or abuse a specific group of people. Often such text is written differently compared to traditional text, such as news involving either explicit mention of abusive words, obfuscated words and typo-logical errors or implicit abuse i.e., indicating or targeting negative stereotypes. Thus, processing this text poses several robustness challenges when we apply natural language processing techniques developed for traditional text. For example, using word- or token-based models to process such text can treat two spelling variants of a word as two different words. Following recent work, we analyze how character, subword and byte pair encoding (BPE) models can aid some of the challenges posed by user-generated text. In our work, we analyze the effectiveness of each of the above techniques and compare and contrast various word decomposition techniques when used in combination with others. We experiment with fine-tuning large pretrained language models, and demonstrate their robustness to domain shift by studying Wikipedia attack, toxicity and Twitter hate-speech datasets.