An Improved Back-Propagation with PSO in MLP-ANNs for Authorship Identification

An Improved Back-Propagation with PSO in MLP-ANNs for Authorship Identification

Azadeh Salamzadeh, Farhad Soleimanian Gharehchopogh

Abstract

Scientific theft and plagiarism (academic papers and books), publishing inappropriate texts, or threatening letters. Therefore, the language is necessary for a security aspect that is recognized by the author. In this paper a larger data set containing 2500 texts for training and 2500 texts for testing are used. A new model for improving Back-Propagation (BP) has been presented with a Particle Swarm Optimization (PSO) algorithm in the Multi-Layer Perceptron Artificial Neural Network (MLP-ANN) for authorship identification of the new model. In this paper, we have used features such as lexical, syntactic and structural features for authorship identification. Finally, different criteria such as the number of correct and incorrect classified data, precision, and recall percentage have been used. The obtained precision and the recall percentage for proposed models are equal to 0.9992 and this criterion in BP is equal to 0.9849 and 0.9588. The above-mentioned results indicate the proposed model is superior to BP.

Keywords

Authorship identification, Natural language processing, Artificial neural networks, Multi-layer perceptron, Particle swarm optimization