Seamless Integration of RESTful Services into the Web of Data

We live in an era of ever-increasing abundance of data. To cope with the information overload we suffer from every single day, more sophisticated methods are required to access, manipulate, and analyze these humongous amounts of data. By embracing the heterogeneity, which is unavoidable at such a scale, and accepting the fact that the data quality and meaning are fuzzy, more adaptable, flexible, and extensible systems can be built. RESTful services combined with Semantic Web technologies could prove to be a viable path to achieve that. Their combination allows data integration on an unprecedented scale and solves some of the problems Web developers are continuously struggling with. This paper introduces a novel approach to create machine-readable descriptions for RESTful services as a first step towards this ambitious goal. It also shows how these descriptions along with an algorithm to translate SPARQL queries to HTTP requests can be used to integrate RESTful services into a global read-write Web of Data. 1. Introduction We live in an era where exabytes of data are produced every single year; never before in human history had we to deal with such an abundance of information. To cope with this information overload, more sophisticated methods are required to access, manipulate, and analyze these humongous amounts of data. Service-oriented architectures (SOAs) built on Web services were a first attempt to address this issue, but the utopian promise of uniform service interface standards, metadata, and universal service registries, in the form of SOAP, WSDL, and UDDI has proven elusive. This and other centralized, registry-based approaches were overwhelmed by the Web’s rate of growth and the lack of a universally accepted classification scheme. In consequence, the usage of SOAP-based services is mainly limited to company-internal systems and to the integration of legacy systems. In practice, however, such a clear and crisp definition of data is rare. Today’s systems integrate data from many sources. The data quality and meaning are fuzzy and the schema, if present, are likely to vary across the different sources. In very large and loosely coupled systems, such as the Internet, the gained adaptability, flexibility, and extensibility, in a transition away from strict and formal typing to simple name/value pairs or triples, outweighs the resulting loss off “correctness.” Thus, it is not surprising that RESTful services, and there especially the ones using the lightweight JavaScript Object Notation (JSON) [1] as the serialization format, are increasingly