Knowledge enhanced schema matching for heterogeneous data integration
Chuangtao Ma · Eötvös Loránd Tudományegyetem · 2023
Knowledge-based integration is regarded as one of the efficient integration methods due to the excellent semantic interoperability of the knowledge base.Ontology-based data integration has played a critical role in heterogeneous data integration and schema mapping because of the superb interoperability and rigorous mathematical foundation of ontology.However, ontology-based data integration and schema matching usually require tailored ontologies, and the traditional methods for constructing ontology are manual, in which a lot of effort and experience from domain experts are required.In particular, the conventional similarity-based schema matching method is incapable of resolving semantic ambiguities and conflicts in some complex mapping cases.To address the aforementioned issues, this dissertation systematically investigates a knowledge-enhanced schema matching for heterogeneous data integration.The goal of this dissertation is to investigate the efficient way to learn and map highquality knowledge bases from the existing relational data models and leverage them to enhance schema matching for heterogeneous data integration.More precisely, it investigates an efficient way to learn and map high-quality ontology from the existing relational data models and leverage these ontologies to enhance and improve conventional schema matching.To begin with, the motivation and research questions are introduced, and then the related works and preliminaries behind knowledge-enhanced schema matching are summarized and given separately.After that, a novel framework for learning and mapping ontology from RDB is designed, the results show that the designed framework of learning ontology from RDB is an alternative method to access the data from various legacy databases and construct the high-quality tailored ontology for schema matching.Additionally, the investigation shows that the potential correspondences v among different attributes of entities in the heterogeneous schema matching and data integration could be easily identified with the help of high-quality tailored ontologies.Aiming for learning high-quality ontologies from RDB, a semi-automatic semantic consistency-checking method is proposed to check the consistency of the learned ontology between its original RDB data model.The verification results show that the model checking and graph-intermediate representation could be employed to construct a (semi-)automatic semantic consistency checking method, which could correctly check and return the results of whether the given semantic specification of the learned ontology satisfies the original RDB model.Moreover, a knowledge-enhanced schema matching method based on knowledge injection and reconciliation is proposed, and a healthcare schema matching task from a real-world scenario is selected as a case study to verify the feasibility and effectiveness of the proposed method.The preliminary experiment shows that there is still room for improvement in the performance of the traditional schema matching method by using the designed knowledge-enhanced schema matching framework, which bridges the gap between entity reconciliation and schema matching for heterogeneous data integration.This dissertation contributes to the semantic web community and database community by exploring the ontology learning framework, consistency checking method, text and knowledge enhancement schema matching for heterogeneous data integration.This work is conducive to handling and resolving the semantic conflicts and ambiguities for complex schema matching in heterogeneous data integration.vi