RDB to Semantic Data Transformation for Semantic Data Publication and Utilization

Abstract

RDB to RDF transformation is a semantic information extraction method that supports the Semantic Web. The direct mapping, one of the RDB to RDF transformation methods, is a representative mapping method recommended by the W3C. The direct mapping processes an automatic mapping from relational data to RDF data. Semantics preservation is an important property of the direct mapping to transform relational data to semantic data without information loss. However, existing direct mapping methods have problems that violate semantics preservation in specific cases. To comply with the semantics preservation, a hierarchical direct mapping method is provided. Rules of the hierarchical direct mapping are defined based on lemmas that represent features of semantic data transformation. A hierarchical semantic vocabulary is also defined to generate sound and precise semantic data. Next, this thesis also focused on developing an effective direct mapping to generate lightweight and intuitive semantic output data. Thus, the optimized hierarchical direct mapping is provided based on a relational meta-schema vocabulary. Rules of multi-column keys are defined to reduce repetitive constraint data generation problems. Rules for multiple keys are also defined because relational tables may contain multiple foreign keys or unique constraints that affect the output data size. The relational meta-schema vocabulary describes concepts of relational data and relationships among the concepts. The optimized hierarchical mapping method uses initially defined relational concepts from the vocabulary, and generates compact and intuitive semantic output data. Finally, a semantic metadata based information retrieval method is provided as semantic data utilization. Existing ranking methods do not have direct methods of evaluating the meaning of links. In this thesis, a semantic metadata based ranking approach is proposed to directly analyze the meaning of links by using a semantic Web data structure. The semantic Web data structure is built upon semantic metadata extracted from the Web data by using the RDB to RDF transformation method described above. The provided method evaluates the weight of the links for stratifying rank values based on their importance in the semantic Web data structure. The experimental results showed that the proposed mapping method performs semantics preserving RDB to RDF transformation and outputs smaller size semantic data with better quality, and the weighted semantic metadata based ranking approach outperforms existing methods.