A Study of Algorithm-Based Detection of Fake News in Brazilian Election: Is BERT the Best
Lara S. Moreira, Gabriel Machado Lunardi, Matheus de Oliveira Ribeiro, Williamson Silva, Fábio Paulo Basso
Abstract
The recent Brazilian election was plagued by the proliferation of false news on the internet. Many people turned to social media to fact-check information and verify its authenticity. In today's digital and data-driven world, fake news can spread rapidly, causing detrimental effects, such as potentially influencing the outcome of an election. In light of this, verifying information has become increasingly reliant on software. While intelligent software can be used to detect and mitigate the spread of fake news, there is a lack of research on the use of such technology in the Portuguese language, particularly when it comes to the implementation of newer strategies such as the Representation of a Bidirectional Transformer Encoder (BERT). Our study evaluated BERT's ability to detect fake news compared to traditional machine learning algorithms, using text classification to identify false news. The results demonstrate BERT's superiority over other algorithms, with a statistically significant difference in all cases. BERT can considered a viable option for detecting fake news.