Managing Data Integration Uncertainty

Synonyms

Probabilistic schema alignment

Definition

Consider a set of source schemas ${\mathcal {S}} = \{S_1, \dots , S_n\}$ in the same domain, where different schemas may describe the domain in different ways. An important component in data integration is schema alignment, including three steps: (1) creating a mediated schema M that provides a unified and virtual view of the disparate sources and captures the salient aspects of the domain being considered, (2) generating attribute matching that matches attributes in each source schema S_i, i ∈ [1, n], to the corresponding attributes in the mediated schema M, and (3) building a schema mapping between each source schema S_i and the mediated schema Mto specify the semantic relationships between the contents of the source and that of the mediated data. The result schema mappings are used to reformulate a user query into a set of queries on the underlying data sources for query answering. Uncertainty can arise in every step of schema...