Automating Misinformation Detection: A Neural Network Approach to News Classification
Disciplines
Artificial Intelligence and Robotics
Abstract (300 words maximum)
The rapid spread of misinformation presents a significant challenge in the digital age, influencing public opinion and shaping societal narratives. This research aims to develop a machine learning model leveraging Natural Language Processing (NLP) to automatically classify news articles based on their misinformation type. The approach consists of three key steps: 1) collecting and utilizing publicly available datasets containing labeled misinformation articles, such as the Fake News Challenge dataset, the LIAR dataset, and Kaggle’s Fake and Real News dataset; 2) fine-tuning pre-trained transformer models like BERT and BART to detect and classify news articles into categories including Clickbait, Satire/Parody, Conspiracy Theories, Biased News, and Objective News; 3) implementing the classification model using TensorFlow and PyTorch to ensure scalability and efficiency. The anticipated outcome is a robust and automated misinformation detection system that enhances users' ability to critically assess news content. This research contributes to ongoing efforts in combating misinformation, with future iterations expanding detection capabilities to multiple languages and social media platforms.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Automating Misinformation Detection: A Neural Network Approach to News Classification
The rapid spread of misinformation presents a significant challenge in the digital age, influencing public opinion and shaping societal narratives. This research aims to develop a machine learning model leveraging Natural Language Processing (NLP) to automatically classify news articles based on their misinformation type. The approach consists of three key steps: 1) collecting and utilizing publicly available datasets containing labeled misinformation articles, such as the Fake News Challenge dataset, the LIAR dataset, and Kaggle’s Fake and Real News dataset; 2) fine-tuning pre-trained transformer models like BERT and BART to detect and classify news articles into categories including Clickbait, Satire/Parody, Conspiracy Theories, Biased News, and Objective News; 3) implementing the classification model using TensorFlow and PyTorch to ensure scalability and efficiency. The anticipated outcome is a robust and automated misinformation detection system that enhances users' ability to critically assess news content. This research contributes to ongoing efforts in combating misinformation, with future iterations expanding detection capabilities to multiple languages and social media platforms.