UC San Diego
Schema mapping for data transformation and integration
- Author(s): Wang, Guilian
- et al.
Finding semantically correct mappings between the schemas of two data sources is a fundamental problem in many important database applications. But it cannot be fully automated, so user input is necessary. This dissertation presents SCIA, a system that assists users in creating executable mappings between a source schema and a target schema by automating simple matches and finding critical points where user input is necessary and maximally useful. SCIA handles complex mappings involving n-to-m matches with semantic functions and conditions, which often exist in practice, but are ignored in most related systems. It outputs mappings in both correspondence format and executable view format that can transform source data into target instances. Formal experiments show SCIA significantly reduces total user effort by using path contexts and combination of multiple matching algorithms to find critical points, and help users at those points by asking them specific questions with sufficient context and suggestions for adding semantic information for data transformation, such as join and grouping conditions. Moreover, SCIA has semantic models for schemas and schema mappings using algebra. This dissertation also gives principles for schema mapping user interface design, based on the ideas of minimal model for schema and schema mapping, and of optimal semiotic morphism from this model into the interface design