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

In recent years, grey social media platforms, those with a loose moderation policy on cyberbullying, have been attracting more users. Recently, data collected from these types of platforms have been used to pre-train word embeddings (social-media-based), yet these word embeddings have not been investigated for social NLP related tasks. In this paper, we carried out a comparative study between social-mediabased and non-social-media-based word embeddings on two social NLP tasks: Detecting cyberbullying and Measuring social bias. Our results show that using social-media-based word embeddings as input features, rather than non-social-media-based embeddings, leads to better cyberbullying detection performance. We also show that some word embeddings are more useful than others for categorizing offensive words. However, we do not find strong evidence that certain word embeddings will necessarily work best when identifying certain categories of cyberbullying within our datasets. Finally, We show even though most of the state-of-the-art bias metrics ranked social-media-based word embeddings as the most socially biased, these results remain inconclusive and further research is required.

Social media have brought threats like cyberbullying, which can lead to stress, anxiety, depression and in some severe cases, suicide attempts. Detecting cyberbullying can help to warn/ block bullies and provide support to victims. However, very few studies have used self-attention-based language models like BERT for cyberbullying detection and they typically only report BERT's performance without examining in depth the reasons for its performance. In this work, we examine the use of BERT for cyberbullying detection on various datasets and attempt to explain its performance by analysing its attention weights and gradient-based feature importance scores for textual and linguistic features. Our results show that attention weights do not correlate with feature importance scores and thus do not explain the model's performance. Additionally, they suggest that BERT relies on syntactical biases in the datasets to assign feature importance scores to class-related words rather than cyberbullying-related linguistic features. CCS CONCEPTS• Computing methodologies → Natural language processing; Supervised learning by classification; • Information systems → Web and social media search.

This paper is a summary of the work in my PhD thesis. In which, I investigate the impact of bias in NLP models on the task of hate speech detection from three perspectives: explainability, offensive stereotyping bias, and fairness. I discuss the main takeaways from my thesis and how they can benefit the broader NLP community. Finally, I discuss important future research directions. The findings of my thesis suggest that bias in NLP models impacts the task of hate speech detection from all three perspectives. And that unless we start incorporating social sciences in studying bias in NLP models, we will not effectively overcome the current limitations of measuring and mitigating bias in NLP models.