How does Semantic Web work?

The technology resolves semantic ambiguities in the descriptions of Web service interfaces by combining information retrieval and semantic Web techniques. Information retrieval techniques are used to resolve the domain-independent relationships. For example, in this approach, semantic similarity is derived using an English thesaurus after “tokenization” and part-of-speech tagging of the names of the elements that describe the interfaces of Web services.

Semantic Web techniques are used to resolve domain-specific similarities. For example, the concepts used in a given domain (such as retail industry, health-care industry, etc.) and the relationships among them are modeled as a domain ontology. The Web services are annotated using semantic annotations from the domain ontologies in Web Services Semantics (WSDL-S) format. Then the ontological similarity of the semantic annotations associated with Web service descriptions is derived by inferring the domain ontology.

Matches from the two approaches are combined to determine an overall similarity score to help assess the quality of a Web service match to a given request. In cases where single services do not match a given request, the system can compose multiple services by employing artificial intelligence (AI) planning algorithms in order to fulfill a given request.

The WSDL-S mechanism for annotating Web services with semantics is based on an approach developed jointly by IBM and the University of Georgia. This approach was submitted to W3C in 2005 and is now the basis for the work of the new Semantic Annotations for WSDL (SAWSDL) Working Group.