Semantic gossiping. Fostering semantic interoperability in peer data management systems
Until recently, most data integration techniques revolved around central approaches, e.g., global schemas, to enable transparent access to heterogeneous databases. However, with the advent of the Internet and the democratization of tools facilitating knowledge elicitation in machine-processable formats, the situation is quickly evolving. One cannot rely on global, centralized schemas anymore as knowledge creation and consumption are getting increasingly dynamic and decentralized. Peer Data Management Systems (PDMS) address this problem by eliminating centralization and instead applying compositions of local, pair-wise mappings to propagate queries among databases. We present a method to foster global semantic interoperability in PDMS settings in a totally decentralized way based on the analysis of the semantic graph linking data sources with pairwise semantic mappings. We describe how quality measures for the mappings can automatically be derived by analyzing transitive closures of mapping operations. The information obtained from these analyses are then used by the peers to route queries in a network of semantically heterogeneous sources, and to iteratively correct erroneous mappings in a self-organizing way. Additionally, we present heuristics to analyze semantic interoperability in large and heterogeneous communities. Finally, we describe Grid- Vine which implements our approach and provides a semantic overlay to demonstrate how our approach can be deployed in a practical setting.