American Research Journal of Computer Science and Information Technology

Vedant Jadhav*1, Shantanu Tripathi1, Yashdeep Shetty1, Aakash Tiwari1 and Arun Chauhan2

The Internet is widely utilized as a platform for exchanging information and ideas. In these encounters, a large volume of textual data is produced. NLP requires a large amount of data for training, which can be found on social networking platforms such as Facebook, Twitter etc.(Shetty) Sarcasm identification, fake news detection, sentiment analysis, and other similar tasks have become possible and valuable. We have trained different transformer-based models on Marathi, a low-resource regional language, and compare their performance in this study (Magueresse et al., 2020). We present that MuRIL gives better performance than other Transformer based models.