The purpose of accessing search engines by most of the users is for answers rather than set of documents. Identifying set of relevant documents for a question is the first part of any question answering systems. The accuracy of a question answering system is highly depending on the quality of the set of documents that have been chosen for further investigation. In this paper, a method is proposed to improve the information retrieval part of the open domain question answering system. The authors found the list of documents that may be relevant to the given question and rank them by using a) alternate queries to increase the quality of the context represented in the query without compromising the semantic meaning, b) search engine results and c) application of the concept of Latent Semantic Analysis (LSA). The result shows that the top ranked documents for a given factoid question contain solid answers. © Technomathematics Research Foundation.