Extracting process patterns from EPC (Event-Driven Process Chain) models requires to perform a semantic analysis of EPC functions and events. An automated semantic analysis faces the problem that an essential part of the EPC semantics is bound to natural language expressions in functions and events with undefined process semantics. The semantic annotation of natural language expressions provides an adequate approach to tackle this problem. This paper introduces a novel approach that enables an automated semantic annotation of EPC functions and events. It employs semantic patterns to analyze the textual structure of natural language expressions and to relate them to instances of a reference ontology. Thus, semantically annotated EPC model elements are input for subsequent semantic analysis.