Competency-based learning has been used in training employees to acquire the necessary skills for an organization to be successful in dynamic and ever-changing environment. One of the core activities in competency-based training is learning material acquisition. Standardization efforts have made the retrieval of educational materials, also called learning objects, easier by describing them in pre-defined metadata schema. However, the existing standardized metadata schema and practices of learning object metadata annotation do not support automatic selection of resources by specific competency requirements in the competency-based learning. We propose an ontology-based competency formalization approach as a way of representing competency-related information together with other metadata in ontology in order to enhance machine automation in resources retrieval. The approach represents competency with properties of definition, knowledge reference, evidence of proficiency, and level of proficiency. The effectiveness of resource selection from each of the property is evaluated.