Analyzing the Impact of BERT Embeddings on Rumor Debunking

Social media allows us to connect and share information with anyone within seconds. Most users access social media through web-based applications on their mobile phones. The ease of information sharing through social media results in the exchange of large amounts of information among its users. This could lead to the spread of rumours, either unintentionally or intentionally. Whichever is the case, rumours can hurt concerned users and sometimes even lead to dire situations. It is impractical to manually filter out rumours from trillions of pieces of information. Thus, various research works have been going on for automatic rumour detection. In this paper, we propose a new way to look at research that has already been done to see how BERT embeddings help find rumours. The proposed model was put through 10-fold crossvalidation. Compared to the baseline implementation, subtask A got better by 3.2% and subtask B got better by 2.9%.