Abstract |
One of the main objectives of the Semantic Web (SW) is to enable interoperability of data published on the Web. The vast majority of these data reside in legacy sources, which are published on the so-called DeepWeb. These sources may vary from relational database systems (with or without XML views) to native XML sources, contain useful information and can be manipulated by a plethora of query languages and searching interfaces. In order to take advantage of these diverse sources we need a SW middleware capable of integrating data residing in these sources by employing rich mediation schemata. In particular, mediation schema should be expressed in terms of a SW language (e.g., RDF/S, OWL) capturing data semantics. The functionality of a SWIM relies on rendering source heterogeneity transparent to the user and facilitating him to query the mediated schema with declarative languages. Additionally, it should support sound and complete algorithms for query reformulation, from the mediated schema to the underlying data sources. In this context, we propose the ICS-FORTH SWIM as a middleware for integrating XML sources by employing domain or application speci¯c RDF/S schemas. SWIM enables users to query the virtual mediated RDF/S schema using a declarative lan- guage (e.g., RQL) and moreover, it provides further levels of abstraction using view de¯nition languages (e.g., RVL). The use of a well-founded fragment of ¯rst-order logic for specifying the mappings between the RDF/S and XML data models exploits the background theory on relational query reformulation and optimization. SWIM adopts a hybrid approach for establishing the mappings between the XML data sources and the mediated RDF/S schema, which is called GLAV.
|