With the advent of internet technology as a ubiquitous platform for sharing the educational contents and experiences, many of the institutions across the globe offer the federated search to the courses, lesson plans, contents, assignments, seminars and experiments. These learning resources are stored in the repositories of the learning content management system. Sophisticated search and information retrieval solutions are essential for efficient use of these repositories. The structure of many existing information retrieval system considers ontology for retrieval. This ontology based solution increases the accuracy of information retrieval through high precision and recall. This paper addresses the requirement for pre-processing and classification of documents in order to achieve more efficient Information Retrieval system. Tools and techniques employed for autonomous classification or clustering of documents are investigated and a new method based on concept expansion is proposed. The proposed methods are evaluated using Reuters 21578 dataset.