Question Answering System with Indic multilingual-BERT

Question answering system is a natural language processing (NLP) task that is very important when it comes to search useful information on websites or large documents. Many language models have been evolved to provide language processing capabilities in various languages. The mBERT is the popular model used to deal with multilingual datasets throughout the world. The research focus was centric to the high resource languages such as English, Hindi, etc. due to the availability of the dataset and many other local languages are overlooked for sentiment analysis. In the present research, a question answering system is focused with Indic Bidirectional Encoder Representations from Transformers (BERT). The performance of Indic BERT with mBERT was analyzed along with the equivalent models.