Fake News Detection in Telugu Language using Transformers Models

In today’s world, lots of people rely on online news every day. But with more people using websites for information, there’s a growing problem of wrong info spreading. This can make it hard to trust news, especially on social media. Detecting fake news online has become really important because it can cause problems for individuals and groups. While there’s been a lot of work done on detecting fake news in popular languages, not much attention has been given to languages with fewer resources. We created a new dataset to address this issue in the detection of fake news in the Telugu language. We used different transformer models like mBERT, XLM-RoBERTa, IndicBERT, and MuRIL for fine-tuning the models in detecting fake news. MuRIL outperformed the rest of these models, obtaining an accuracy of 88.79%. MuRIL demonstrated the highest accuracy in classifying more number misleading news correctly.