Content-based publish/subscribe systems as a cooperation and integration middleware are gaining increasing popularity due to its loose coupling and scalable integration capability. Recently Web-based collaboration develops very fast. RDF as a standard for information exchange in the semantic web and RDF Site Summary (RSS) based on RDF is becoming popular on the web, however the RSS systems based on pull architecture do not scale well because the websites hosting the popular RSS feeds face high loads imposed by active polling and clients don?t detect updates in time. It needs a novel architecture as publish/subscribe system to improve its scalability and efficiency. However existing publish/subscribe systems only have the simple subscription mechanism, and lack of supporting the RDF data. In the paper we introduce a scalable semantic publish/subscribe system for efficient dissemination of the RSS documents. Our system adopts RDF and OWL ontology language to understand the semantic of events. We also propose a highly efficient event matching algorithm based on RDF graph. The prototype of SPS (Semantic Publish/subscribe System) was developed to selectively disseminate the RSS documents, and the performance tests demonstrated its efficiency and scalability.