Research On Ontology Semantics Based Natural Language Watermarking Technology

With the rapid development of Internet, copyright disputes raised by the spread of multimedia digital text are becoming an imperative problem. In order to protect the author's authentication, text digital watermarking technology came into exist.Traditional text watermarking technology is almost based on the modification of text format to embed watermarking which could not be easily sensed by human. For example, the space between word and line, the color and size of characters could all be exploited for embedding. Such kind of method did have some effects. However the biggest shortcoming is the robustness. The attacker could remove the watermarking by copying the contents to notepad with no other efforts.In order to overcome the weakness of traditional watermarking system, a new method based on synonym substitution named Equmark is proposed. It made the watermarking more concealed. But such method could not prevent the attacker covering the original watermarking by substituting the word randomly.Later on, researcher decided to move to a higher lexical level to launch watermarking research. Atallah proposed a scheme of transforming sentence structure to embed information. Its main idea is to let some bits of sentence exactly match the watermarking by transforming the sentence structure. This kind of method increased the watermark capacity and could be implemented easily. After that, a few demos came into existence, such as Enigmark and so on.In recent years, the hot area of nature language watermarking is turned to semantically-based watermarking. Semantically-based watermarking introduced the TMR tree to represent the meaning of a sentence. It lets some bits of the TMR tree exactly match the watermarking by transforming the TMR tree. It has a wide range of transforming forms and gained more flexibility and larger watermark capacity.Around the facts listed above, this paper does the following work:1,Proposed L semantically watermarking scheme. L scheme mainly improved the robustness of original semantically-based watermarking by introducing TMR resort preprocessing mechanism.2,Proposed L watermark threshold. The experiments under such threshold could achieve the highest ratio of detecting.3,Proposed a watermark recover strategy by the idea of longest common strings.At last, a demo system has been implemented. This prototype provides 9 TMR samples to test the efficiency of L scheme. It verifies the best threshold of L scheme by repeatedly comparison on 3 different thresholds. All these experiments show that L watermarking scheme is effective.