Domain-specific web resource discovery is a key issue for resources sharing and reuse. Considering the text semantic information inside enterprise sites reflects the website structures and textual content feature, a novel website topic feature modeling method are presented in this paper. The website classification algorithm on the basis of the hybrid vector employs the effective kind of support vector machine (SVM). Manufacturing enterprise website search from the web is implemented to verify the classification performances. Moreover, the accuracy and efficiency of enterprise sites search are improved by revised search strategies. Finally, a prototype system of web enterprise resource service is introduced. It preliminarily demonstrates that this method is of good applicability for specific web resource discovery.