Around the globe, we are witnessing a transition from a primarily goods-based to a service-based economy. At the same time, advances in information technology provide significant opportunities for the electronic provisioning of services via the internet. As a result, more and more services are being traded online. Just as search tools are indispensable to discover information or goods on the web, they will become vital for discovering services on the web. In this paper we present Themis-S, a prototype of an ontology-based natural language service discovery engine. In a series of experiments, we evaluate the retrieval effectiveness of Themis-S in comparison to other state-of-the-art information retrieval models. The experiments indicate that Themis-S, incorporating WordNet as a general purpose ontology, can outperform systems applying syntactic information retrieval models such as the Vector Space Model (VSM) or the Probabilistic Relevance Model (PRM).