Towards a Holistic Schema Matching Approach Designed for Large-Scale Schemas

Holistic schema matching is a fundamental challenge in the big data integration domain. Ideally, clusters of semantically corresponding elements are created and are updated as more schemas are matched. Developing a high-quality holistic schema matching approach is critical for two main reasons. First, identifying as many accurate and holistic semantic correspondences as possible right from the beginning. Second, reducing considerably the search space. Nevertheless, this problem is challenging since overlapping schema elements are not available. Identifying schema overlaps is further complicated for two main reasons: (1) there is a large number of schemas; and (2) overlaps vary for different schemas. In this paper we present HMO, a Holistic schema Matching approach based on schema Overlaps and designed for large-scale schemas. HMO can balance the search space and the quality of the holistic semantic correspondences. To narrow down the search space, HMO matches schemas based on their overlaps. To obtain high-accuracy, HMO uses an existing high-quality semantic similarity measure. Experimental results on four real-world domains show effectiveness and scalability of our matching approach. © 2020, Springer Nature Switzerland AG.