This paper focuses on the generation of domain ontology. At first, the paper introduces the traditional search engines and pointed out their shortcomings, in addition, it also illustrates that retrieval should be based on the semantic web, Then the concepts of semantic web, domain ontology and so on are proposed, next it makes a research in the plsa algorithm of extracting domain concepts and the k-means algorithm of clustering those concepts, finally, it shows a football ontology constructed by protégé, and makes a prospect to semantic retrieval based on ontology.