The automatic discovery and composition of Web services rely on the facilities offered by the providers of services in describing the functionalities of their services semantically and on the description of the client's requests and their needs, without being ambiguous. In this paper, we present a model of Semantic Annotations for Web Services Discovery and its Composition. The proposed approach uses an inter-connected network of semantic Web services describing in OWL-S, using the similarity measure (outputs-inputs similarity) between concepts based on ontology, built before any submitted request. In only one exploration, the composition algorithm can find several composition plans. But the selected composition plan must be "the best one " according to the quality criteria (similarity, time and memory space). This technique takes advantages from a graph structure, chaining algorithm of expert system and semantic annotations.