LLMs enhance hate speech detection by generating contextual data · arXiv research

Researchers have developed methods to improve hate speech detection by using Large Language Models (LLMs) to generate contextual information for social media posts. These methods involve incorporating the generated context into a SBERT-based classifier through techniques like text concatenation, embedding concatenation, and hierarchical transformer fusion. Evaluations on the Latent Hatred dataset for textual hate speech and the MAMI dataset for multimodal hate speech showed performance improvements of up to 3 and 6 F1 points, respectively, compared to a zero-context baseline. AI

IMPACT Improves accuracy of AI systems designed to detect and mitigate online hate speech.

RANK_REASON Academic paper detailing a new methodology for hate speech detection using LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• Joshua Brookes
• Latent Hatred dataset
• LLMs
• MAMI dataset
• SBERT

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →