SeMap: A Generic Schema Matching System

The rapidly growing number of autonomous data sources on the web makes the need of effective tools of creating semantic mappings increasingly crucial. Moreover, the goal of allowing applications to have more expressive semantics requires a change in focus. While most previous work focus on creating mappings in specific data models for data transformation, they fail to capture a richer set of possible relationships between schema elements. For example, current schema matching approaches might discover that ‘TA’ in one schema equals to ‘grad TA’ in another one, even though the relationship can be modeled more accurately by saying that ‘grad TA’ is a specialization of ‘TA’. This increased semantics of the mapping in turn allows for applications involving richer semantics. In this thesis we concentrate on the following problem: given initial match (correspondence) information produced by current schema matching techniques, how to construct a complex, semantically richer mapping that can be used across data models? Specifically, we aim at detecting the relationship types of ‘Has-a’, ‘Is-a’, ‘Associates’ and ‘Equivalent ’. Technically, we achieve this goal in mainly three steps: (1) exploiting various types of semantic evidence for possible matches; (2) finding a globally optimal match assignment; (3) identifying the relationship embedded in the selected matches. We implemented our semantic matching approach within a prototype system SeMap, and tested its accuracy and effectiveness.