LLM in the Loop: Creating the ParaDeHate Dataset for Hate Speech Detoxification
This addresses the problem of limited detoxification data for researchers and practitioners, offering a scalable alternative to human annotation, though it is incremental as it builds on existing ParaDetox methods.
The paper tackles the scarcity of high-quality parallel datasets for hate speech detoxification by proposing an LLM-in-the-loop pipeline using GPT-4o-mini to create ParaDeHate, a dataset of over 8K hate/non-hate text pairs, and shows that models fine-tuned on it achieve better performance in style accuracy, content preservation, and fluency.
Detoxification, the task of rewriting harmful language into non-toxic text, has become increasingly important amid the growing prevalence of toxic content online. However, high-quality parallel datasets for detoxification, especially for hate speech, remain scarce due to the cost and sensitivity of human annotation. In this paper, we propose a novel LLM-in-the-loop pipeline leveraging GPT-4o-mini for automated detoxification. We first replicate the ParaDetox pipeline by replacing human annotators with an LLM and show that the LLM performs comparably to human annotation. Building on this, we construct ParaDeHate, a large-scale parallel dataset specifically for hatespeech detoxification. We release ParaDeHate as a benchmark of over 8K hate/non-hate text pairs and evaluate a wide range of baseline methods. Experimental results show that models such as BART, fine-tuned on ParaDeHate, achieve better performance in style accuracy, content preservation, and fluency, demonstrating the effectiveness of LLM-generated detoxification text as a scalable alternative to human annotation.