Designing and refining ontologies becomes a tedious task, once the boundary to real-world-size knowledge bases has been crossed. Hence semi-automatic methods supporting those tasks will determine the future success of ontologies in practice. In this paper we describe a way for ontology creation and refinement by combining techniques from natural language processing (NLP) and formal concept analysis (FCA). We point out how synergy between those two fields can be established thereby overcoming each other’s shortcomings.