We investigate Web service composition as a planning problem and use the input-output parameter relations in order to select the constituent services that make up the composite service. Furthermore, we make use of ontological information between the input-output parameters such that a more specific concept can be used instead of a general concept to make the process more flexible. Our proposed approach is based on constructing a dependency graph including the service parameters and Web services themselves. By using this dependency graph, we perform backward chaining starting to search from the desired output parameters, which is in fact the goal, to the available input parameters. In addition to using semantic information through the search, our approach considers non-functional attributes of the services such as service quality. Considering the quality measures, we find the constituent services by making use of depth first search. After finding the required services, our algorithm generates a plan that shows the execution order of each service.