Abstract: Most of information retrieval and text summarization methods does not describing about the semantics of events and these methods rely only on shallow document representation. The current problem of Information retrieval is that query given by the user is not the same as the one by which the information has been indexed. It is exceptionally hard to locate the required data and relevant document. Hence, structuring the queries and documents in terms of event graph using supervised machine learning and rule based model and employ graph kernels for query document similarity.