基于不变特征的自然语言鲁棒水印技术（英文版）

英文标题：Robust Natural Language Watermarking through Invariant Features中文摘要：本文研究一种基于自然语言的水印系统，旨在通过嵌入足够的信息和提取不变特征以提高其负载能力和鲁棒性，进而提高水印系统的版权保护能力。通过系统分析提出了一种抗干扰的填充模型，并建立了一个对比实验，结果表明该方法相较于之前的方法鲁棒性平均提高了百分之十六点八。英文摘要：Recent years have witnessed a proliferation of valuable original naturallanguage contents found in subscription-based media outlets, web novelplatforms, and outputs of large language models. Without proper securitymeasures, however, these contents are susceptible to illegal piracy andpotential