Satire or Fake News? Machine Learning-Based Approaches to Resolve the Dilemma

The rise of the internet and social media platforms has led to a significant increase in the spread of fake news in recent years. This rapid dissemination, especially in sensitive areas such as politics and finance, has caused considerable harm, prompting researchers to explore ways of automatically identifying misinformation through language structure. However, distinguishing between fake news and satirical content remains a challenge due to the nuanced nature of satire. While much research has focused on detecting fake news, relatively little attention has been given to the specific task of differentiating it from satirical news. To bridge this gap, this study investigates the performance of nine traditional AI models and three trans-former models named BERT, XLM-RoBERTa, and DistilBERT in distinguishing among fake and satirical news. Initially, a Support Vector Machine (SVM) classifier shows superior performance on a few data, particularly when data preprocessing techniques like text-stemming are applied. But, after employing augmentation techniques, the transformer based models can surpass the traditional methods, achieving an accuracy of 98 %. This suggests that with a larger, more diverse dataset spanning different topics and timeframes, we can reliably differentiate between both news.