CANARDEEP: A MODEL FOR RUMOUR DETECTION IN BENCHMARK PHEME DATASET

Abstract

A rumour is any statement that is not yet confirmed at the time of posting, irrespective of whether it’s true or false. It is evident that rumours are an imperious threat to the credibility of the information providers. The sheer volume of information diffusion has led to an imperative need for questioning the tangibility of information. Unsubstantiated rumours on social media can cause significant damage by deceiving and misleading the society. It is essential to develop models that can detect rumours and curtail its cascading effect and virality. In this project, we proffer a CanarDeep model for rumour detection in the benchmark PHEME dataset. The proposed model is a hybrid deep neural model that combines the predictions of a hierarchical attention network (HAN) and a multi-layer perceptron (MLP) learned using context-based (text + meta-features) and user-based features respectively. A logical OR based decision-level late fusion strategy is used to dynamically combine the predictions of both the classifiers and output the final label as rumour or non-rumour. The results validate superior classification performance to the state-of-the-art. The model can facilitate timely intervention by buzzing an alarm to the moderators and further forming a cordon to inhibit the dissemination of spurious and junk content.