The task of teaching a machine to go through a passage and answer a question, like a human would is called Machine Reading Comprehension (MRC). It is a new and challenging topic in Machine Learning. This project uses the technique of Bi-directional Attention in Deep Neural Networks, originally put forward by the people of Allen Institute for Artificial Intelligence, University of Washington. Previous models using Attention, would use it to summarise the passage into a fixed size vector, or take only Uni-directional Attention of Query on the Context. The advantages of this technique, along with the pipeline of this model are explored in great detail. The model is trained on the SQuAD dataset, and implemented using TensorFlow (Version 2.3).

[Note: Code is present in section 4.12 of the report]