Search

We created a fine-grained AI system for the detection of anti-Semitism. This Explainable AI will identify English and German anti-Semitic expressions of dehumanization, verbal aggression and conspiracies in online social media messages across platforms, to support high-level decision making., 6 pages

This PhD thesis investigates and analyses the effectiveness of text classification models for detecting toxic language in paediatric settings. The literature highlights toxic language, and its effects like bullying and mental health problems, as fundamental societal challenges. Moreover, the World Health Organisation asserts that tackling bullying for adolescents should not be limited to educational settings and that it is the responsibility of healthcare institutions to address these issues. Social media platforms have implemented text classification systems that protect against toxic language within their products, and paediatric wards should have comparable safeguards when using language-based technology. The thesis aims to expose methods from Natural Language Processing that are suitable for application in paediatrics, and highlight aspects of state-of-the-art methodology that demand consideration and attention. The thesis is structured in three parts; an introduction, a series of case studies, and a strategic analysis. The introduction is targeted at non-expert readers and intends to support the technical case studies chapters by clarifying systems and practices from the field of Natural Language Processing. The case studies part contains a series of contained experiments in text classification and toxic language detection. The final part returns to the systems from the case studies and analyses them against the context of paediatric application.