Large-sized repositories of learning objects (LOs) are difficult to create and also to maintain. In this paper we propose a way to reduce this drawback by improving the classification mechanisms of the LO repositories. Specifically, we present a solution to automate the LO classification of the Universia repository, a collection of more than 15 million of LOs represented according to the IEEE LOM standard. Although a small part of these LOs is correctly classified, most are unclassified and therefore searching and accessing information is difficult. Our solution makes use of the categories provided by DBpedia, a linked data repository, to automatically improve its classification through a graph-based filtering algorithm, which selects the most suitable categories for describing a LO. Once selected, these categories will classify the LO, linking its classification metadata with a set of DBpedia categories. (Contains 6 tables and 6 figures.)