Modeling Profanity and Hate Speech in Social Media with Semantic Subspaces
This addresses data sparsity issues in hate speech detection for multiple languages, including German, English, French, and Arabic, but is incremental as it builds on existing representation methods.
The paper tackled the problem of hate speech and profanity detection in social media, which suffers from data sparsity and annotation incompatibility, by identifying semantic subspaces in representations and testing them in zero-shot settings across languages; it resulted in improvements of F1 scores between +10.9 and +42.9 over baselines.
Hate speech and profanity detection suffer from data sparsity, especially for languages other than English, due to the subjective nature of the tasks and the resulting annotation incompatibility of existing corpora. In this study, we identify profane subspaces in word and sentence representations and explore their generalization capability on a variety of similar and distant target tasks in a zero-shot setting. This is done monolingually (German) and cross-lingually to closely-related (English), distantly-related (French) and non-related (Arabic) tasks. We observe that, on both similar and distant target tasks and across all languages, the subspace-based representations transfer more effectively than standard BERT representations in the zero-shot setting, with improvements between F1 +10.9 and F1 +42.9 over the baselines across all tested monolingual and cross-lingual scenarios.