Classification of Fake News by Fine-tuning Deep Bidirectional Transformers based Language Model

• Main Authors: Akshay Aggarwal, Aniruddha Chauhan, Deepika Kumar, Mamta Mittal, Sharad Verma
• Format: Article
• Language: English
• Published: European Alliance for Innovation (EAI) 2020-10-01
• Series: EAI Endorsed Transactions on Scalable Information Systems
• Subjects: fake news; transfer learning; deep learning; natural language processing
• Online Access: https://eudl.eu/pdf/10.4108/eai.13-7-2018.163973

With the ever-increasing rate of information dissemination and absorption, “Fake News” has become a real menace. Peoplethese days often fall prey to fake news that is in line with their perception. Checking the authenticity of news articlesmanually is a time-consuming and laborious task, thus, giving rise to the requirement for automated computational tools thatcan provide insights about degree of fake ness for news articles. In this paper, a Natural Language Processing (NLP) basedmechanism is proposed to combat this challenge of classifying news articles as either fake or real. Transfer learning on theBidirectional Encoder Representations from Transformers (BERT) language model has been applied for this task. This paperdemonstrates how even with minimal text pre-processing, the fine-tuned BERT model is robust enough to performsignificantly well on the downstream task of classification of news articles. In addition, LSTM and Gradient Boosted Treemodels have been built to perform the task and comparative results are provided for all three models. Fine-tuned BERTmodel could achieve an accuracy of 97.021% on NewsFN data and is able to outperform the other two models byapproximately eight percent.