Introducing Semantic Annotations

The World Wide Web is a vast collection of arbitrary resources, held together by standards that tell us how to locate, transport and visualize content. This obviously includes textual information on a Web site, but we are more interested in actual data such as satellite images, Web services providing vector data, or real-time sensor data. Without proper descriptions, the use of such content is limited. Before publishing it to the Web, it has to be annotated with descriptive metadata to make it usable to a broad audience. Otherwise people will neither be able to find the resource using search engines nor will they be able to evaluate if the discovered resource satisfies their current information need.

Existing metadata standards don't tell us much about what the served data (or process) represents, and in particular they lack a way to link the resources to external models. For example, an application may know how to load and visualize the data on a map, but a user has no idea how to read the displayed map. With the help of semantic annotations, data providers will be able to connect the standardized service descriptions to the modeled knowledge. Such models comprise conceptualized knowledge about the represented geographic phenomena. Having such a link established, reasoning algorithms will be able to infer if a Web Service matches an agent's query on the formal level. In addition it will allow for extracting valuable contextual information from the knowledge models, making it possible to display thematic information for the displayed data and helping the user to understand.

A depth description of the semantic conflicts which can arise if two actors don't agree on the interpretation of the content is available in the discussion paper. Once we have a reference to an external semantic descriptions established, these semantic conflicts can be solved using, for example, a reasoning engine such as  Pellet or  IRIS. The annotation approach is independent from the logic used for the external knowledge models. It can either be a fully specified formal ontology expressed using OWL, or a simple controlled vocabulary served, for example, using SKOS.

In our notion, a semantic annotation is a reference from some arbitrary structured (encoded in XML or XML-Schema) content, to a knowledge model which re-models the content's model in another dialect and extends it with contextual information. The  W3C standard "Semantic Annotations for WSDL and XML Schema" (SAWSDL) introduces the modelReference as element which holds this link, but is restricted to WSDL Web service descriptions and XML-Schema documents. Sapience supports SAWSDL, but extends its application to other standards, including XML entities. We understand the model reference to be the link between different encodings. In addition, we introduce the domain reference as link between local, application-specific documents and the globally shared and commonly accepted knowledge models.