The article offers methods and means of mild duplicate text fragment search on the base of semantic text network analysis. Extracting of semantic network of text as the text semantic portrait is their base. The network is preparing with technology of automatic semantic text analysis Text-Analyst, and then is using for texts sense comparing. Proposed approach uses several levels of texts sifting for fast and exact text duplicates revealing: for the first time fast semantic networks comparison and then — revealing of mild copies of text fragments.