Section: Scientific Foundations
Semantic interoperability ensures that the meaning of the information that is exchanged is automatically interpreted by the receiver of a message. In centralized systems, this property improves the relevance of query answers. In distributed heterogeneous systems, it is compulsory to enable autonomous heterogeneous sources understand each other to obtain relevant results.
To provide semantic interoperability within a system, much research has been conducted on semantic representations. The main idea is to use meta-information which eases the meaning understanding. This approach needs the definition of ontologies which describe the concepts and relations between them, for a given domain. During the last fifteen years, much effort has focused on formal methods to describe ontologies, ressource description languages, reasoning engines...All these methods represent the fundations of the semantic web. However, many works rely on the assumption that a single ontology is shared by all the participants of the system.
However, in distributed systems with autonomous participants, such as P2P systems, this assumption is not realistic anymore. On the contrary, one has to consider that the participants create their ontologies independently of each other. Thus, most often the ontologies differ. To tackle this problem, research on ontology matching proposes several techniques to define corresondances between entities of two ontologies. So, in some way, ontology matching highlights the shared parts of two ontologies. Thus it provides the basis for interoperability between heterogeneous participants and by “transitivity” in the whole system.
Although ontology matching and other semantic web techniques provide a basis for interoperability, the challenge is still to define a whole semantic infrastructure in which participants' search for information is both relevant and efficient. Considering semantics can be useful at different stages. First, semantic representation of queries and information may improve the relevance of the results. It can be used in place or in addition to usual request representation. Second, semantics can be used to represent participants, or groups of them, leading participants to better know each other. Such information can be useful for routing the requests to other participants in order to obtain the relevant answers within a short time and with a low trafic load. Third, this information can also be used to organize the network so as to improve efficiency. All these research directions have received partial answers but more work is needed on the interaction between all these elements and their impact on the efficiency of the global system.