A Schema Matching Method Based on Partial Functional Dependencies

Schema matching is a difficulty in many database application domains, e.g., data integration, E-business, data warehousing and semantic query processing. We can get correct schema mapping by mining the semantics of elements from the elementspsila own information (e.g., elementspsila names and elementspsila data types and domains), data instances and structure information. But in fact, most existing related works only consider elementspsila own information, and data instances and structure information are seldom used for schema matching. At present, there is a trend to combine the elementpsilas own information with elementspsila data instance information and structure information for schema matching in order to improve the matching accuracy. The new method proposed in this paper uses elementspsila data instances and structure information to support matching. For a pair of element x and y, if a very small number of tuples are deleted from the table, x fully functionally determines y. Such kind of functional dependencies are called partial functional dependencies. A set of strategies are introduced in this paper which utilize these partial functional dependencies to improve schema matching efficiency and accuracy. Extensive simulation experiments are conducted and the results show that this method is better than other related methods in various performance metrics such as precision, recall and overall.