Title: Multi Document Reading Comprehension

Abstract: Reading Comprehension (RC) is a task of answering a question from a given passage or a set of passages. In the case of multiple passages, the task is to find the best possible answer to the question. Recent trials and experiments in the field of Natural Language Processing (NLP) have proved that machines can be provided with the ability to not only process the text in the passage and understand its meaning to answer the question from the passage, but also can surpass the Human Performance on many datasets such as Standford's Question Answering Dataset (SQuAD). This paper presents a study on Reading Comprehension and its evolution in Natural Language Processing over the past few decades. We shall also study how the task of Single Document Reading Comprehension acts as a building block for our Multi-Document Reading Comprehension System. In the latter half of the paper, we'll be studying about a recently proposed model for Multi-Document Reading Comprehension - RE3QA that is comprised of a Reader, Retriever, and a Re-ranker based network to fetch the best possible answer from a given set of passages.