Text Imbalance Handling and Classification for Cross- platform Cyber-crime Detection using Deep Learning
Munipalle Sai Nikhila, Aman Bhalla, Pradeep Singh · 2020
Cyberbullying has become a very prevalent issue in recent times. Not just qualifying this as a women issue, starting from politicians to a ten-year-old kid, every person is being bullied on any social platform. It is highly essential to build an artificial intelligence based model that detects the presence of bullying in cross-platform posts. However, textual datasets which are useful for model generation are highly imbalanced in nature. In this paper, we propose two main methods to handle textual data imbalancing which are synonym replacement and artificial data generation using generative adversarial neural networks. We present a systematic analysis of our approaches using a convolutional neural network classifier. Our work shows how removing data imbalance with generative adversarial network techniques before classification improves the overall performance of the model.