This paper addresses the challenge of open-domain question answering using Wikipedia as the sole knowledge source. The authors propose a system called DrQA, which combines a document retriever and a document reader to efficiently find and answer questions from Wikipedia. The document retriever uses bigram hashing and TF-IDF matching to identify relevant articles, while the document reader is a multi-layer recurrent neural network trained to detect answer spans within those articles. Experiments on multiple QA datasets show that both components outperform existing methods, and multitask learning with distant supervision significantly improves overall performance. DrQA demonstrates strong capabilities in answering factoid questions from Wikipedia, outperforming existing systems and achieving state-of-the-art results on the SQuAD benchmark.This paper addresses the challenge of open-domain question answering using Wikipedia as the sole knowledge source. The authors propose a system called DrQA, which combines a document retriever and a document reader to efficiently find and answer questions from Wikipedia. The document retriever uses bigram hashing and TF-IDF matching to identify relevant articles, while the document reader is a multi-layer recurrent neural network trained to detect answer spans within those articles. Experiments on multiple QA datasets show that both components outperform existing methods, and multitask learning with distant supervision significantly improves overall performance. DrQA demonstrates strong capabilities in answering factoid questions from Wikipedia, outperforming existing systems and achieving state-of-the-art results on the SQuAD benchmark.