Firstly, the extraction of precise concept and their relationship, and the attributes of precise concept are completed according to the related terminology standard. Secondly, Protégé is used to create and maintain rough ontology, so that OWL rough ontology can be stored into a relational database in structured manner [13]. Thirdly, Jena API is used to complete relation reasoning between concepts, attributes of related concepts are obtained, and then semantic similarity can be calculated. Fourthly, after the annotation of the document retrieval resources made by semi-automatic annotation tool, the index database is created by using IndexWriter in Lucene, and then the matching and ranking of the extended concepts set and the annotated documents can be performed, that is, semantic matching based on Jena and semantic retrieval are finished. Finally, the result is returned to the user. Moreover, the software development platform and the query interaction interface of ROSRS are built by Java and Eclipse, on which the proposed semantic information model in Fig. 2 is realized.