On Estimating Semantic Loss in Peer Data Management Systems

Traditional data integration techniques have focused on centralized components to allow transparent querying of multiple heterogeneous databases. This requirement on global/mediated schemas impedes the deployment and success of such systems in modern highly dynamic and scalable computing environments. Peer Data Management Systems (PDMS) offer a truly decentralized solution by utilizing compositions of local pair-wise mappings to propagate queries through the network. Current PDMS proposals have been ignorant of the fact that query reformulation could result in semantic degradation due to the inherent heterogeneity of the nodes. A critical issue to effectively address this problem is the existence of a technique that computes the query difference, which results from this reformulation. In this work, we propose an algorithm that estimates the semantic loss of the rewritten queries in a generic way, based on the notion of containment mapping. This information can then be used as the basis for extending appropriately the schema mappings and improving the quality of the retrieved answers.