This paper presents a requirement ontology which can represent both sentence level semantics which means semantics in the structure of a sentence, and word level semantics which means the meaning of a word. In this way, complete semantics of a sentence can be described. The Generalized Upper Model (GUM) is a widely used linguistically motivated ontology which supplies a hierarchy of general concepts for defining objects, actions and properties. WordNet is a worldwide used lexical database for English language. In the requirement ontology, structures of a sentence are described as relations between sentence components which are derived from GUM. And meanings of key words are explained by WordNet. Thus, semantics of natural language requirements (NLRs) are captured for further processing.