EXCHANGING DATA FROM INSTITUTIONAL REPOSITORIES TO THE SEMANTIC WEB

Abstract

Internet and semantic web technologies have enabled academics to find online research materials with increasing speed and accuracy. They have enabled academics to make connections with each other. Whereas, institutional repositories (IRs) are often built to serve a specific institution’s community of users. Mostly existing IRs are using relational database schema for maintaining the metadata of their digital contents. They might need to interact with other information systems that build to manage institutional research activities. Thus, it is crucial to provide interoperability and integration mechanisms to bridge the gap between the semantic web and relational database worlds. To process the data in semantic context, a relational database is transformed into ontology. The use of semantic web technologies in integrating the different IRs metadata enable ontology-facilitated sharing and reuse of learning resources. They provide users access to a web of content which might otherwise require discovering and exploring multiple websites or IRs. The main promising feature of IRs is their flexible data models that can be customized to arrange the digital documents in a repository according to the organizational structure of an institute. The data model of an organization’s IR is not directly converted into IR database schema, but the data model schema is maintained as values in the comprehensive database schema of the IR. The schema of IRs databases is nested schema i.e. a schema is embedded in another schema. In other words, an IR database schema is not a normalized schema with respect to the data model, so, it makes the transformation complicated and different from the typical transformation tasks. A substantial amount of research has already been done to transform a relational database into ontology. However, these systems are only capable to transform a normalized relational database into ontology. They cannot produce accurate results if they are applied on IR databases. After building the ontologies, a key issue is to enable interoperability among different ontologies. vii viii The proposed system first of all identifies the data model of an institute from IR database and builds a normalized relational schema for the data model of the institute. Then metadata of the repository is extracted to populate this produced schema to build an intermediate database. Once we get a normalized relational database, then relational to ontology transformation techniques are applied on this intermediate database to transform it into ontology. After that, the system transforms the instances from the generated ontology into corresponding data or instances expressed in target ontology. The classes from both source and target ontologies are extracted and simple mappings between these classes are generated by the user. Then the individuals of these mapped classes are matched and proper URIs are given to each individual. These individuals are linked with their respective target ontology classes. Finally, an RDF, having individuals of the target ontology, is generated. The system has mainly three modules: (i) Metadata Extraction; (ii) Relation to Ontology Transformation; (iii) Ontology Alignment and Data Translation. The distinguishing features of the proposed system are (i) identifying the data model of an IR; (ii) extracting metadata of the repository; (iii) creating proper hierarchy of parent and child classes of ontology to preserve the data model hierarchy, (iv) generating mappings between ontologies, and (v) transforming data or instances from source ontology into corresponding data or instances expressed in target ontology. The system has been implemented in Java language and Jena API is used for ontology creation. Experimental results demonstrate that the transformation is correct and the system preserves information capacity.