Comparative Analysis of Various Text Summarization Techniques via Leveraging Transformer Model for the Fake News Detection

Abstract

Today, the internet has come to be a necessary part of our lifestyle. The role of traditional information channels consisting of newspapers and televisions on how we acquire and consume news has to become much less prominent than within the past. Indeed, the boom of social media structures has performed a critical function in this variation. Oppositely, it empowers the widespread of ”fake news,” i.e., low-quality news with purposefully false data. The broad spread of fake news has the potential for incredibly adverse effects on people and society. The proposed research work aims to design a robust model for an automatic fake news detection system to help journalists and everyday users from misleading content. In this paper, we have studied and performed a deep comparison of leading Transformer-based models for the task of text classification and explored and compared various text summarizing techniques for dealing with more prominent long-length articles before classifying them through existing models. BERTSUM gives the most noticeable results out of all the three methods by enabling us to create a system to label an arbitrarily long article as fake or genuine