Cyberbullying text detection in social media based on a deep learning model

Abstract

Cyberbullying in social media is becoming more and more common because of the rapid development in the communication domain. The close connection between people brings the bullying manner from the physical world into cyberspace and does harm to the network environment. This thesis aims to utilize deep learning techniques to detect cyberbullying text. Based on two open-source datasets, we apply the pre-trained GloVe model to obtain the word embedding. We employ the RCNN model to classify the cyberbullying text with TensorFlow. During training, this thesis compares the different performance under different model configs. We show the process of optimization and get a well trained model. At last, we make a conclusion regarding the work we have done. Besides, we propose several directions for future work on coping cyberbullying manner.