Peering into the Mind of Language Models: An Approach for Attribution in Contextual Question Answering