Abstract: A robust text-watermarking scheme based on significant components is presented. The text significant components are acquired by analyzing text feature. According to watermarking sequence and the evaluation model of synonymy substitution, the watermark is imperceptibly embedded into the significant components of the text, which improves the robustness of watermarking. Furthermore, majority voting is used to improve the robustness of watermarking and reduce the probability of false positives. Theoretical analysis and experimental results have indicated that the proposed scheme can improve the robustness, security, and invisibility of watermarking.