Extracting Digital Fingerprints from Chinese Documents

It is an important problem to extract features from Chi- nese documents for protecting intellectual property. The ex- isting approaches are major oriented to words frequency or semantic, they can't extract features efficiently. By mapping Chinese documents into an ordered set of integers, we find that a Chinese document can be corresponded to a unique ordered set of integers and the set is an isomorphism of the document. So, we propose an algorithm which can hash the set to three kinds of hash value sequences: paragraph se- quence, sentence sequence and chunk sequence, which can represent the features of the document completely. In or- der to reduce the numbers of the features defined as digital fingerprints in this paper, we present an optimal strategy to select some hash values from the sequences. The experiment results show that the algorithms proposed are efficient.