Racism Detection in Twitter Using Deep Learning and Text Mining Techniques for the Arabic Language

In light of the recent cultural withholding of discriminative tendencies, cyber-racism detection is no longer luxurious incorporation to online platforms. As a result, the detection of hate speech has been a vital area of research in the last decade. Nevertheless, the progress of development and incorporation of hate speech detectors in general and racism detectors in particular in the Arab region is considered to be relatively slow in comparison to English speaking countries. Therefore, this work aims to contribute to the detection of cyber-racism in the Arab region by considering the complicated nature of the Arabic language through tailored data acquisition and multi-staged pre-processing models. Furthermore, it takes into account the specificity of the application of racism detection in comparison to more broad sentiment analysis models that validly call for complicated architectures. Our proposed approach utilizes deep learning techniques due to their superiority given the availability of very large datasets. In detail, it employs both convolutional neural network (CNN) and metaheuristic optimization algorithm for the classification of racism-oriented tweets in the Twitter platform.