📝 Tox-BART: Leveraging Toxicity Attributes for Explanation Generation of Implicit Hate Speech
👥 Authors: Neemesh Yadav, Sarah Masud, Vikram Goyal, Vikram Goyal, Md Shad Akhtar, Tanmoy Chakraborty
📌 Access the full paper here: https://arxiv.org/abs/2406.03953

In this video cutting-edge research that employs language models to generate clear explanations for implicit hate posts has been presented. This work aims to reveal the underlying stereotypes and support content moderators in their crucial efforts. Despite the common approach of integrating top-k relevant knowledge graph (KG) tuples to enhance understanding, our findings challenge the effectiveness of this method.

🔍 What's Covered:
The process of using language models to identify and explain implicit hate
Analysis of the impact of knowledge graph quality on explanation generation
Surprising results showing simpler models with external toxicity signals outperform KG-based approaches
Discussion on leveraging toxicity attributes over KG infusion with Pre-trained Language Models (PLMs)
Explore these intriguing findings and understand why shifting away from knowledge graphs might be the future of combating hate speech online.



Subscribe to stay informed about the latest advancements in AI and digital content moderation!