A Holistic Approach to Schema Matching

Schema matching, which establishes whether two objects are semantically related, has been a focus of data management research due to its critical role in enterprise information integration. Current schema matching techniques typically tend to exploit only a single method selected from matching schema structure, or syntax, or semantics, or data and probability distributions, which reduces the success of schema matching. An alternative approach to improve the effectiveness of schema matching is to use a holistic approach that integrates and generalizes techniques based on structural and syntactic comparisons, as well as probability distribution and mutual information. This paper describes this holistic approach and its implementation, and presents initial results of comparisons with existing techniques.