Text Copy Detection Research On Fingerprint Feature

The fingerprint feature-based text copy detection approach can quickly identify the plagiarization.but suffers from some weak points,such as the fingerprint feature is too large,complexity process of fingerprint feature extraction and low performance in the similarity calculation.To resolve the problems in fingerprint feature extraction,we proposed two algorithms,namely mixed window-based fingerprint feature extract method and optimal decision-based fingerprint feature extract method.Mix window-based fingerprint feature extraction method is based on the idea of fuzzy hashing algorithm,in which the fingerprint feature is extracted according to the trigger point.In our case,we combine the fixed window and sliding window to select the trigger point for fingerprint feature extraction.Optimal decision-based fingerprint feature extract method is based on winnowing algorithm,introduces the theory of optimal decision,and constructs the model of optimal decision for the fingerprint feature extraction.To resolve the inefficiency of similarity calculating,we proposed an improved edit distance-based similarity calculation algorithm.The algorithm predicts the sub-sequence between the comparing fingerprint features and redefines the similarity formula in the process of calculating edit distance.Experimental results show our proposed method can reduce the amount of the fingerprint feature,and improve the efficiency of the fingerprint feature extraction.Improved edit method can enhance the applicability and speedy of similarity formula.