Knowledge-Grounded Target Group Language Recognition in Hate Speech

Hate speech comes in different forms depending on the communities targeted, often based factors like gender, sexuality, race, or religion. Detecting it online is challenging because existing systems are not accounting for diversity of hate identity target and may be biased towards certain groups, leading to inaccurate results. Current language models perform well identifying communities, but only provide a probability that text contains references particular group. This lack transparency problematic these learn biases from data annotated by individuals who familiar with To improve detection, particularly group identification, we propose new hybrid approach incorporates explicit knowledge about used specific groups. We leverage Knowledge Graph (KG) adapt it, considering an appropriate level abstraction, recognise speech-language related gender sexual orientation. A thorough quantitative qualitative evaluation demonstrates our as effective state-of-the-art while adjusting better domain changes. By grounding task knowledge, can contextualise results generated proposed groups most frequently impacted technologies. Semantic enrichment helps us examine model outcomes training detection systems, handle ambiguous cases human annotations more effectively. Overall, infusing semantic crucial enhancing understanding behaviors addressing derived data.