While the NLP literature has recently seen one groundbreaking technique after another in feature representation and embedding, there is still work that to be done in imparting knowledge into embeddings, beyond the use of contextual information; moreover, most of the techniques require huge computing time and power, while training from scratch. The medical field is an area where feature representations that are interpretable, and that need limited computing resources to design, can benefit the adoption and acceptance of these systems in practice. Through the OrderSketch algorithm presented in this paper, we aim to provide a step towards a knowledge-rich feature representation technique that is based directly on ontologies and is also computation-resource friendly. The paper is organized as follows: First we present related work on other knowledge representation and order detection techniques. Next we introduce our algorithm OrderSketch, which is simple and yet also effective in capturing knowledge and order information from ontologies. We apply our embedding algorithm to WordNet, and to an augmented medical ontology called SnoMeSHNet that we introduce. After presenting our current constructions and experiments, we discuss future experiments and goals for this work, in progress.