Offensive Span

Offensive span detection focuses on identifying the specific words or phrases within a text that contribute to its overall offensiveness, going beyond simple sentence-level classification. Current research emphasizes multilingual capabilities, leveraging transformer-based models like BERT and RoBERTa, often incorporating attention mechanisms to pinpoint offensive spans and their targets. This work is crucial for improving online content moderation and fostering healthier online environments by enabling more nuanced and accurate identification of harmful language across diverse languages and platforms.