Geospatial Web Services need to adopt Semantic Web technologies in order to become more clearly and machine-readably understandable to the wider Web, as well as to make accessible the many large and diverse geodatasets coming on line at a frightening pace. OWL-S and technologies for working with it show promise; however there is a need to enable both the processing of geospatial relationships and the description of tightly coupled service content ,in order for Geospatial Semantic Web Services to become a reality.
The geospatial information community moved early into the Web Services world, with the aim of enabling even the simplest browser application and least expert user to leverage the capabilities of immense data repositories and processing facilities being deployed across the globe. In this process new and rigorous requirements were generated for both Web Services and general IT technologies. The need to begin connecting disparate information communities as well as reach out effectively to wider non-specialized audiences, has subsequently led organizations such as the Open Geospatial Consortium (OGC) and its members to investigate Semantic Web Services technologies. The rub is that they also have baggage to bring here. This includes specialized ontologies; specialized processing of geospatial rules and relationships; and a concept of Web Services being both self-describing and often tightly coupled to highly dynamic geospatial content. We are confident; however, that frank dialogue between the Spatial Web and the Semantic Web communites in venues such as workshops and testbeds, as well as practical experimentation such as described below can result in a Geospatial Semantic Web framework which both can live with.
The Interoperability Experiment (IE) is a methodology developed by OGC to provide a means of testing the effect of new technologies, new architectures, and new applications on existing or planned OGC specifications.
The GSWIE will address several important steps towards the development of a Geospatial Semantic Web (GSW), where discovery, query, and consumption of geospatial content are to be based on formal semantic specification. The experiment seeks specifically to test the extension and implementation of Web Feature / Filter Encoding (WFS / FE) services where these may be combined with existing semantic tools to implement a relatively simple, but distributed and end-to-end geospatial query use case. Significant elements of this experiment include:
Development and encoding of formal geospatial ontologies, including feature type descriptions, which leverage existing OGC OWS Services markup, Geographic Markup Language (GML) , and metadata standards such as FGDC, ISO 19115, 19119, etc.;
Geospatial service interfaces which can provide service information formulated in the OWL-S semantic expression language referencing those ontologies;
WFS / FE interfaces which can operate on requests for ontologically expressed service and content descriptions;
Semantic query language processing interfaces for WFS /FE which operate on combinations of the above ontologies, service information and queries;
Tools for generating semantically expressed geospatial information;
Implemented WFS / FE components for use in processing geospatial queries
While several deliverables are described for this Interoperability Experiment, the most important deliverables will be the test demonstration of an end-to-end semantic geospatial query itself, where WFS /FE data and services relevant to the query are discovered through knowledgebase reasoning, and change proposals for the WFS / FE specifications which reflect the results of the experiment.
Sample Geospatial (Intelligence) Query
Specifically challenging concepts here for ontology processing include “within 500 miles” as a relationship, “Kandahar” as a place, and “support” as a reasoning rule in combination with “C5A aircraft” and “airfields”.
One of the most specific advantages which Geospatial Web Services give is global access to geodatasets which are created and kept up to date by particular organizations at particular installations, obviating the need for myriad facility, distribution, and update issues with traditional means of data distribution. “Maintain locally, access globally” runs somewhat counter to the usual methods of knowledgebase processing, which largely assume that all knowledge potentially relevant to a given problem can be fetched and assembled in one contiguous graph store.
The “Service-Oriented Architecture” (SOA) approach to this problem in the existing OGC Services world has been to utilize artfully generalized descriptions of distributed service and data resources, available either from a self-description service interface (GetCapabilities) or from a separate Catalog Service, in order to maximize the chances that a particular service request will return (part of) the answer to the question at hand. While these descriptions have in the past been peculiar to the OGC services and encodings which use them, there is increasing use of wider standards, as well as an increasing effort to leverage WSDL for at least an alternative generalized service description.
In the SOA “Publish-Find-Bind” paradigm, however, these descriptions are used for both finding and for binding, in fact they are being used for for “finding the right binding”, both in terms of appropriate service (process) binding and in terms of relevant content binding.
As the GSWIE looks at ways to preserve the distribution efficiencies of the present Spatial Web and still leverage knowledgebase processing of ontologies, we are looking at ways to implement virtual knowledgebases (presently within the SESAME framework) which are “lazily loaded” when the descriptions of remote services indicate that particular service request are likely to yield additional knowledge relevant to the query being processed.
The markup language OWL-S provides many of the concepts and constructions which would be useful for this task. The service profile component of OWL-S provides a first level of discovery as to a remote service's relevance, while the grounding and process components can form the basis for a virtual knowledgebase plug-in (a SAIL in SESAME terminology). The affinity between OWL-S and WSDL also provides a more tractable transition from the existing service descriptions to formal semantic ones.
As described earlier, there are specific requirements of a Geospatial Semantic Web which place additional demands on service description markup such as OWL-S. A critical need is to discover in some detail the content which is currently bound to a particular service. The difficulty of keeping such information up to date for a highly dynamic service, such as a satellite sensor or georeferenced video service is one of the reasons that OGC services have been specified with their own catalog operation. This could certainly be one approach to using OWL-S, defining additional self-description operations for the content of a service.
What an OGC GetCapabilities operation returns, however, is a complete service description. New self-description operations would still be self-referential, in that they would need to provide not only the description of content, but the particular binding to content, in other words the content profile and grounding to compliment the more static operations descriptions.
We propose, therefore, that OWL-S be refined to accommodate such content binding so that its markup can be better used to “find the binding” and better support distributed knowledgebase processing of geospatial queries.
The Semantic Web provides exciting challenges and opportunities for the Spatial Web, and we feel that the inverse is also true. Refinements to knowledgebase processing and Semantic Web Service discovery will help the Geospatial Semantic Web to succeed, but can also help the wider Semantic Web Services framework to be stronger and more versatile. We hope that the open exchange of ideas, technologies, and expertise can make this so.