Gender Identification: A Comparative Study of Deep Learning Architectures

摘要

Author profiling, dating back to the earliest attempts at of analyzing quantitative text documents, is an extensivel-studied problem among NLP researchers. Because of its utility in crime, marketing and business. In this paper, three deep learning methods were evaluated for author profiling using tweets in Arabic language. The first method is based on a Convolutional Neural Network (CNN) model, while the second and third technique belongs to the family of Recurrent Neural Networks (RNN). The appropriate choice of some parameters, such as the number of amount of filters, training epochs, batch size, dropout and learning rate of Adam optimizer used in a RNN model is crucial in obtaining reliable results. The experimental findings of our comparative evaluation study demonstrate that GRU model outperforms LSTM and CNN models.