Abstract
Many extant natural language watermarking techniques demand deep structure analysis, and so suffer in reliability. We propose a scheme for natural language watermarking, which embedding watermark bits into the pragmatics feature of text by rewriting sentences. In contrast, we eschew syntactic and semantic analysis. We make use of transformation templates and our templates based on pragmatics rule and described by part-of-speech (POS) tags order. The searching structure of pragmatics in the text is simplified as matching the POS tags order with templates. Since there lack pragmatics parser and we proposed a spread spectrum coding scheme, our method resist against synonymic substitution, adverbs inclusion or removal, passivization, topicalization, extraposition, preposing, etc. The experiment results show that no more than 5% watermark bits had been damaged while 30% sentences were transformed without changing the meaning.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
