Early question-answering (QA) approaches could only adequately answer fact-related questions and are insufficient in answering complex questions. More sophisticated QA systems utilize the semantics of words to answer complex questions. However, to answer certain kinds of complex questions, it is necessary to perform deep reasoning by utilizing the semantics of words as well as domain knowledge such as the effects and executability of actions. We propose a QA system that uses AnsProlog to represent domain knowledge and reason from the knowledge extracted by using Link Grammar and WordNet. We use the travel domain to illustrate our approach.