Tox-BART: Leveraging Toxicity Attributes for Explanation Generation of Implicit Hate Speech

TL;DR

This paper introduces Tox-BART, a model for explaining implicit hate speech that leverages toxicity signals rather than knowledge graphs, showing simpler models can outperform KG-based approaches in explanation quality.

Contribution

Proposes Tox-BART, a novel approach that uses toxicity attributes for explanation generation, challenging the effectiveness of knowledge graph integration in this task.

Findings

Simpler toxicity-based models outperform KG-infused models.

Tox-BART achieves comparable or better performance on standard metrics.

Human evaluation indicates more precise explanations than GPT-3.5 zero-shot.

Abstract

Employing language models to generate explanations for an incoming implicit hate post is an active area of research. The explanation is intended to make explicit the underlying stereotype and aid content moderators. The training often combines top-k relevant knowledge graph (KG) tuples to provide world knowledge and improve performance on standard metrics. Interestingly, our study presents conflicting evidence for the role of the quality of KG tuples in generating implicit explanations. Consequently, simpler models incorporating external toxicity signals outperform KG-infused models. Compared to the KG-based setup, we observe a comparable performance for SBIC (LatentHatred) datasets with a performance variation of +0.44 (+0.49), +1.83 (-1.56), and -4.59 (+0.77) in BLEU, ROUGE-L, and BERTScore. Further human evaluation and error analysis reveal that our proposed setup produces more precise explanations than zero-shot GPT-3.5, highlighting the intricate nature of the task.