This thesis proposes a query expansion mechanism based on ontology. Automatic query expansion has facilitated web pages search in several ways. An external knowledge resource can help user identify the searching domain and efficient keywords. Ontology is taken as the metadata of a knowledge domain. Query could be expanding in different approaches based on ontology. In this research, an ontology learning process is implemented. With no initial ontology as backbone, domain ontology is constructed from World Wild Web document semi-automatically. Three expanding approaches based on concepts and their relations are proposed. Ontology learning result and expanding approaches are evaluated by comparing the different search results in atypical IR system.