On the basis of analysis and study on the open source Lucene system architecture, a semantic search system is designed based on the special XML data sources in this paper. What's more, we use the word item location and word semantic to improve the Lucene's search results and conduct experiments to test and verify the retrieval performance, the accuracy of similarity search, the space efficiency of index and the time-efficiency of supporting inquiry: And finally by deploying the Tomcat server to implement our implement system. The experiment results prove that compared with the original Lucene indexing system, our system can improve the indexing efficiency, query efficiency and accuracy.