Abstract:

In this paper a novel technique is proposed to generate hashing values fortext documents. The approach uses Recurrent Neural Networks (RNNs) for this purpose.RNNs are dynamic and temporal type of Neural Networks (NNs) that evolve continuouslybased on subsequent vectors of inputs. The capabilities of RNNs to incorporate presentvalues of inputs with previous values exploiting relations and semantics of the text makeit a competitive paradigm to discover the internal representations within text data in aunique way. Two types of RNNs are tested and compared to traditional methods. Ade-quate review has been done to existing techniques and the results obtained in this workdemonstrate the applicability of this arti cial intelligence paradigm in generating hashingvalues for plain text. RNNs are highly exible, compact, and parallel in nature. Theircapabilities are exploited in this paper as future competent technique in text hashing.Keywords: Recurrent neural network, Hashing methods, Collision probabilities, Intelligentparadigms, Message digest