Author Identification Based on Convolutional Neural Network

In this study, we present an author identification model utilizing the capabilities of Convolutional Neural Networks (CNN). The primary focus of the study is on finding the optimal network structure, where the size of the extracted feature vectors by CNN is monitored in relation to the achieved accuracy. Additionally, the research examines the network structure and complexity such as the number of convolutional layers, their filters, the size and strides of pooling, etc. The proposed model demonstrates high effectiveness in identifying authors achieving an accuracy of 81.99% on test data classifying 50 authors. The model was trained and evaluated using the Reuters 50–50 dataset. These results highlight the potential of CNNs in author identification tasks.