Top-K Generation of Mediated Schemas over Multiple Data Sources

Schema integration has been widely used in many database applications, such as Data Warehousing, Life Science and Ontology Merging. Though schema integration has been intensively studied in recent yeas, it is still a challenging issue, because it is almost impossible to find the perfect target schema. An automatic method to schema integration, which explores multiple possible integrated schemas over a set of source schemas from the same domain, is proposed in this paper. Firstly, the concept graph is introduced to represent the source schemas at a higher-level of abstraction. Secondly, we divide the similarity between concepts into intervals to generate three merging strategies for schemas. Finally, we design a novel top-

k

ranking algorithm for the automatic generation of the best candidate mediated schemas. The key component of our algorithm is the pruning technique which uses the ordered buffer and the threshold to filter out the candidates. The extensive experimental studies show that our algorithm is effective and runs in polynomial time.