Breaking the Cloak! Unveiling Chinese Cloaked Toxicity with Homophone Graph and Toxic Lexicon
C$^2$TU is a novel, training-free method for identifying hidden toxic words in Chinese text that are disguised using homophones or similar sounds. It works by first finding potential toxic words using a graph of Chinese homophones and a known toxic word list. Then, it filters these candidates using either BERT or LLMs to correct the cloaks into their actual toxic meanings, achieving superior results on Chinese toxicity datasets.