TY - BOOK
T1 - THE MANIPULATION OF SCHEMATIC CORRESPONDENCES WITH THE QUANTIFICATION OF UNCERTAINTY IN DATASPACES
AU - Mao, Lu
PY - 2013
Y1 - 2013
N2 - Dataspaces aim to remove upfront cost in the generation of the schema mappingsthat reconcile schematic heterogeneities, and to incrementally improve thegenerated mappings based on user feedback. The reconciliation of schematic heterogeneitiesis a crucial step for translating queries between a mediating schemaand data sources. The generation of schema mappings depends on the elicitationof conceptually equivalent schema constructs and information on schematic heterogeneities.Furthermore, many dataspace operations manipulate associationsbetween schemas, for example for generating a global schema to mediate userqueries. With a view to minimizing upfront costs associated with understandingthe relationships between schemas, many schema matching algorithms and toolshave been developed for postulating equivalent schema constructs. However, theyderive simple associations between schema constructs, and do not provide richinformation on schematic heterogeneities. Without manual refinement, the elicitationof conceptually equivalent schema constructs and schematic heterogeneitiesmay create uncertainties that must be managed.The schematic correspondences captures a wide range of one-to-one and manyto-many schematic heterogeneities. This thesis investigates the use of schematiccorrespondences as a central component in a dataspace management system. Tosupport query evaluation in a dataspace in which relationships between schemasare represented using schematic correspondences, we propose a mechanism forautomatically generating schema mappings from the schematic correspondences.We then characterise model management operators, which can underpin the bootstrapingand maintenance of dataspaces, over schematic correspondences. Tosupport the management of uncertainty in dataspaces, we propose techniques forquantifying uncertainty in the equivalence of schema constructs from evidence inthe form of similarity scores and user feedback, and provide a flexible frameworkfor incrementally updating the uncertainties in the light of new evidence.
AB - Dataspaces aim to remove upfront cost in the generation of the schema mappingsthat reconcile schematic heterogeneities, and to incrementally improve thegenerated mappings based on user feedback. The reconciliation of schematic heterogeneitiesis a crucial step for translating queries between a mediating schemaand data sources. The generation of schema mappings depends on the elicitationof conceptually equivalent schema constructs and information on schematic heterogeneities.Furthermore, many dataspace operations manipulate associationsbetween schemas, for example for generating a global schema to mediate userqueries. With a view to minimizing upfront costs associated with understandingthe relationships between schemas, many schema matching algorithms and toolshave been developed for postulating equivalent schema constructs. However, theyderive simple associations between schema constructs, and do not provide richinformation on schematic heterogeneities. Without manual refinement, the elicitationof conceptually equivalent schema constructs and schematic heterogeneitiesmay create uncertainties that must be managed.The schematic correspondences captures a wide range of one-to-one and manyto-many schematic heterogeneities. This thesis investigates the use of schematiccorrespondences as a central component in a dataspace management system. Tosupport query evaluation in a dataspace in which relationships between schemasare represented using schematic correspondences, we propose a mechanism forautomatically generating schema mappings from the schematic correspondences.We then characterise model management operators, which can underpin the bootstrapingand maintenance of dataspaces, over schematic correspondences. Tosupport the management of uncertainty in dataspaces, we propose techniques forquantifying uncertainty in the equivalence of schema constructs from evidence inthe form of similarity scores and user feedback, and provide a flexible frameworkfor incrementally updating the uncertainties in the light of new evidence.
M3 - Doctoral Thesis
PB - University of Manchester
ER -