Although there has been a surge of interest in applying domain ontologies to facilitate communications among computers and human users, engineering of these ontologies turns out to be very labor intensive and time consuming. Recently, some learning methods have been proposed for automatic or semi-automatic extraction of ontologies. Nevertheless, the accuracy and computational efficiency of these methods should be improved to support large scale ontology extraction for real-world applications. This paper illustrates a novel domain ontology extraction method. In particular, contextual information of the knowledge sources is exploited for the extraction of high quality domain ontologies. By combining lexico-syntactic and statistical learning approaches, the accuracy and the computational efficiency of the extraction process can be improved. Empirical studies have confirmed that the proposed method can extract reliable domain ontology to improve the performance of information retrieval and facilitate human users to discover and refine domain ontology.