SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators
This work addresses the scarcity of parallel multilingual datasets for text detoxification, which is a problem for researchers and developers working on multilingual text processing tasks.
The authors tackled the problem of multilingual text detoxification by introducing SynthDetoxM, a dataset of 16,000 multilingual parallel text detoxification sentence pairs, and achieved superior performance to existing datasets, with models trained on SynthDetoxM outperforming all evaluated LLMs in few-shot setting.
Existing approaches to multilingual text detoxification are hampered by the scarcity of parallel multilingual datasets. In this work, we introduce a pipeline for the generation of multilingual parallel detoxification data. We also introduce SynthDetoxM, a manually collected and synthetically generated multilingual parallel text detoxification dataset comprising 16,000 high-quality detoxification sentence pairs across German, French, Spanish and Russian. The data was sourced from different toxicity evaluation datasets and then rewritten with nine modern open-source LLMs in few-shot setting. Our experiments demonstrate that models trained on the produced synthetic datasets have superior performance to those trained on the human-annotated MultiParaDetox dataset even in data limited setting. Models trained on SynthDetoxM outperform all evaluated LLMs in few-shot setting. We release our dataset and code to help further research in multilingual text detoxification.