Research Of Chinese Text Zero-watermark Based On Statistical Characteristics

At the present, a large amount of digital texts spread on the Internet, which could easily be nondestructively copied and unauthorizedly used, facing the increasing copyright problems. So, how to protect these texts’ copyright has become a key research in the field of copyright protection.After analyzed and summarized the existing research results about copyright protection for Chinese texts, taking content and semantic characteristics of texts into consideration, we designed and constructed a zero-watermark scheme based on Chinese texts’ statistical characteristics using related technologies of computational linguistics. We discussed each part of the scheme in detail, and our main work and study are as follows:(1) We analyzed and summarized the digital watermark technology and its application in text copyright protection, and then designed the zero-watermark scheme based on Chinese texts’ statistical characteristics. This scheme contains three parts:text pretreatment, zero watermark construction and zero watermark detection.(2) In the part of text pretreatment, we discussed part-of-speech tagging and the basic method of dividing a text into sentences or words, designed and implemented some key algorithms, such as generating candidate word set and getting sentence feature information.(3) In the part of zero watermark construction, we gave the total process flow, and designed some key algorithms in detail, such as extracting feature words, calculating average length of sentences and short sentence rate, generating sentence feature vector and POS statistical characteristics, fusing and scrambling the watermark, etc.(4) In the part of zero watermark detection, judged the detected text’s copyright ownership by the way of weighting sum with multi-feature similarities, and designed some key algorithms in detail, such as parsing and restoring the stored watermark, calculating watermark similarity, judging text’s copyright ownership and so on.(5) Discussed and assessed the way of determining candidate word set and the selection of key parameters, analyzed our scheme’s performance and performed some experiments to verify it. The results showed that it has better effectiveness and availability.