It is proposed to establish inverted indexing for ontology-annotated outputs when services are registered in order to find the target services in a quick, accurate and efficient way. For each ontology-annotated output, there is a service list which records all the services in the registry that deliver the output. Based on the indexing, a composition-oriented service discovery algorithm is proposed, which greatly accelerates the filtering of irrelevant atomic services by making use of the inverted indexing, and increases the likelihood of finding a possible candidate by exploring service composition. The results of the extensive experiments show that the proposed algorithm provides better performance on response time than the sequential matchmaking, and better recall rate than the algorithms without the exploration of composition.