This paper presents an information extraction system which can precisely extract semantic elements from texts by using OWL-based domain ontologies. In the system, each inputted text is viewed as a non-formalized instance of a domain ontology. And the outputted semantic elements extracted from the text are formalized elements composed of classes, individuals and properties. To perform the extraction, two extracting algorithms are proposed and used in the system. They are "semantic information extracting algorithm" and "semantic information re-recognizing algorithmrdquo. When the extraction begins, ldquosemantic information extracting algorithm" will extract classes, individuals and explicit properties from the text and store them in the database. After that, the text will be scanned by the "semantic information re-recognizing algorithm" and some useful information that is not found by "semantic information extracting algorithm" will be extracted. Experiments show that the results are accurate with two algorithms working cooperatively.