An empirical study on optimizing query transformation on semantic peer-to-peer networks

Ontology mapping is critical for semantic interoperability between information systems in ontology-based distributed environments. Manual ontology mapping by human experts has been studied as traditional approach. However, these manual tasks are usually expensive, so that it is difficult to obtain mapping results between all possible pairs in a large-scale distributed information system. Thereby, in this paper, we propose a system to estimate the ontology mappings in an indirect manner by making the existing mappings collaboratively sharable and exchangeable, and more importantly, efficiently composing the collected existing mappings. In particular, this work focuses on query propagation for searching for relevant resources on the distributed networks. Once indirect mapping from source system to destination is obtained, the queries can be efficiently transformed to automatically exchange knowledge between them by referring to the mappings, even though they do not have direct connection. In order to evaluate the proposed mapping composition method, we have measured the ratio (i.e., precision and recall) of the indirect mappings to reference mappings which were acquired from human experts. It means that we have regarded information loss by query transformation as an important indicator to knowledge sharing in ontology-based distributed environment.