The recent publications in the field of online abuse detection and moderation highlight a significant shift towards leveraging advanced language models (LMs) and large language models (LLMs) for more nuanced and effective detection mechanisms. These technologies are being employed to tackle a wide range of issues, from hate speech and cyberbullying to more specific forms of abuse like Hinduphobia and the identification of online sexual predators. A notable trend is the emphasis on creating and utilizing high-quality, culturally annotated datasets to train these models, ensuring that detection systems are sensitive to the socio-cultural contexts in which abuse occurs. Furthermore, there is a growing recognition of the dual role of LMs and LLMs in both mitigating and exacerbating online abuse, prompting a call for ethical considerations and safeguards in their deployment.

Noteworthy Papers

• A survey of textual cyber abuse detection using cutting-edge language models and large language models: Offers a comprehensive analysis of online abuse forms and the dual role of LMs and LLMs in detection and generation of abusive content.
• HP-BERT: A framework for longitudinal study of Hinduphobia on social media via LLMs: Introduces a novel framework for analyzing Hinduphobic discourse, highlighting the correlation between COVID-19 case spikes and increases in Hinduphobic rhetoric.
• AfriHate: A Multilingual Collection of Hate Speech and Abusive Language Datasets for African Languages: Presents a significant contribution to the field with a multilingual dataset for African languages, emphasizing the importance of local context in hate speech detection.
• A Survey on Pedophile Attribution Techniques for Online Platforms: Reviews current methods for identifying online sexual predators, pointing out the lack of effective attribution tools and listing open research problems.