Exposing relational data on the Semantic Web with CROSS

With the advent of the World Wide Web, a growing amount of heterogeneous data sources is becoming available, thus requiring wrapping mechanisms for a unified access by end-users or applications. OWL, an ontology language for the Web, is a good candidate for coping with structural and semantic heterogeneity between data sources. For this purpose, we propose CROSS, an open-source prototype which aims at exposing relational databases and additional knowledge about them in OWL. The key feature of CROSS is that it follows a declarative approach since it relies on the target language, OWL. In this approach, the relational structure is first converted into OWL which can then be employed by the user to declare additional OWL statements about the database. The conversion rules are inspired by previous work on database reverse engineering. However, the originality of our declarative approach is that those rules are used without any exception; we show in the paper with a number of examples how additional knowledge can be added without ever questioning the rules. We also discuss a number of directions for further work, such as the choice of the target language and the integration of the CROSS prototype with the Sesame framework.