NLP-driven Content Classification Towards Fake News and Bully Detection

Fake news and cyberbullies proliferate in today’s society, and detecting them at their onset is paramount because of their anti-social elements that affect people’s life and the trust-worthiness of online platforms, particularly social media, like Twitter. However, it is a herculean task due to the unstructured nature of natural language and the limited availability of curated datasets to build automated solutions. Recent studies show that deep learning (DL) inspired natural language processing (NLP)-driven solutions can overcome the challenges. In this direction, this study introduces a lightweight convolutional neural network (CNN) with an attention mechanism to address the fake news and cyberbully detection problems. A thorough experimental study conducted on multiple benchmark datasets proves that the proposed model achieves extremely competitive results. It records significant improvements of 1% -35% when compared to existing baseline models with respect to the benchmark datasets.