A Multi-Linguistic Fake news detector on Hindi, Marathi and Telugu

Nowadays the use of the internet to read news is becoming much more common than our conventional way of reading newspapers, due to this fake news articles are increasing on the internet and also humans are breaking the barrier of languages through translation tools, which gives rise to the fake contents. There are various NLP models available to detect fake news, but these fake news detectors are designed for the English language only, and according to Twitter MD: Non-English tweets are 50% of total tweets, and we don't have language-specific detectors for all languages. So this is where we come up with our Multi Linguistic Fake news detection model. Our model is trained using concepts of transfer learning because it is difficult to get a large amount of data for local or Indic languages as compared to English; we are using a bert-base-multilingual-uncased version of BERT provided by Google. Our Dataset is coming from popular checker portals: BuzzFeed and PolitiFact.