Safeguarding text generation API's intellectual property through meaning-preserving lexical watermarks

Shiyu Zhu, Yun Li, Xiaoye Ouyang, Xiaocheng Hu, Jipeng Qiang

Abstract: We aim to protect text generation APIs in this work. Previous LW methods compromised text quality and made watermarks easy to detect through error analysis due to not considering polysemy. To fit this, we propose meaning-preserving lexical substitution method that considers the target word’s correct meaning in context x. This enables high-confidence identification while making watermarks more invisible.