Hate Speech Annotation and Automatic Assessment: Resource and Best Practices Development

In the wake of the research I have undertaken for my Master's degree's thesis on automatic detection and assessment of right-wing extremists' online speech, I intend to leverage the insights gained thereby to help CLiPS refine and expand their resources in the field of hate speech. It will include, among other possible sub-tasks, the automatic collection of hate speech textual data, the definition of formal annotation guidelines for such data building on existing literature such as "Hate Speech Dataset from a White Supremacy Forum" (De Gibert et al., 2018), the fine-grained annotation of the data for ulterior automatic processing, investigate the best automatic assessment methods for the data (both in terms of performance and explainability), study the opportunities and limitations (from a technical, ethical and practical perspective) of said automatic processing, transpose the results to develop hate speech assessment tools for other languages such as French or German.