Efficient fuzzy search is an information retrieval task, which aims at extracting a condensed version of the original term in document. In this paper, we propose a context sensitive retrieval approach which rearranges terms in effective manner. A new approach using the lexical association between terms to give a context sensitive weight to the document terms has been proposed. The proposed sentence similarity measure has been used with the baseline graph-based ranking models for sentence extraction. It is always difficult to find relevant information in unstructured text documents. In this paper we study the methods of fuzzy search, instant search and proximity ranking and how they can be used in the process of annotation of documents. These various methods can be integrated to give better search results and to achieve efficient space and time complexities. We propose a novel alternative approach which facilitates the generation of the structured metadata automatically using OpenNLP, methods of Instant-fuzzy search and Proximity ranking. It is done by identifying documents which are likely to contain the information of interest. And this information will be subsequently useful for querying the database. Instant search is an emerging information retrieval platform in which a system can find the answers to a search query instantly while the user is typing and before user hits the search button. When user start typing character in search text box, at the same time the system start finding the relevant answers to the query as user progress typing. Fuzzy search helps to improve user search experience by suggesting the user possible search keyword and correcting the typographical mistakes.