Automatic Service Composition is a research direction aimed at facilitating the usage of atomic web services. Particularly, the goal is to build workflows of services that solve specific queries, which cannot be resolved by any single service from a known repository. Each of these services is described independently by their providers that can have no interaction with each other, therefore some common standards have been developed, such as WSDL, BPEL, OWL-S. Our proposal is to use such standards together with JSON-LD to model a next level of semantics, mainly based on binary relations between parameters of services. Services relate to a public ontology to describe their functionality. Binary relations can be specified between input and/or output parameters in service definition. The ontology includes some relations and inference rules that help to deduce new relations between parameters of services. To our knowledge, it is for the first time that parameters are matched not only based on their type, but on a more meaningful semantic context considering such type of relations. This enables the automation of a large part of the reasoning that a human person would do when manually building a composition. Moreover, the proposed model and the composition algorithm can work with multiple objects of the same type, a fundamental feature that was not possible before. We believe that the poor model expressiveness is what is keeping service composition from reaching large-scale application in practice.