Designing a knowledge-based schema matching system for schema mapping

Schema mapping that provides a unified view to the users is necessary to manage schema heterogeneity among different data sources. Schema matching is a required task for schema mapping that finds semantic correspondences between entity pairs of schemas. Semi-automatic schema matching systems were developed to overcome manual works for schema mapping. However, such approaches require a high manual effort for selecting the best combinations of matchers and also for evaluating the generated mappings. In order to avoid such manual works, we propose a Knowledge-based Schema Matching System (KSMS) that performs schema mapping both at the element and structure level matching. At the element level matching, the system combines different matching algorithms using a hybrid approach that consists of machine learning and knowledge engineering approaches. At the structure level matching, the system considers hierarchical structure that represents different contexts of a shared entity. The system can update knowledge if schema data changes over time. It also gives facilities to the users to verify and validate the schema matching results by incremental knowledge acquisition approach where rules are not predefined. Our experimental evaluation demonstrates that our system is able to improve the performance and to generate the accurate results.