ONTOLOGY DEVELOPMENT IN THE SEMANTIC WEB FOR EARTH AND ENVIRONMENTAL TERMINOLOGY (SWEET)

Web searches for Earth science data and information are commonly hindered by syntax mismatches between information user and information provider. If the user does not enter the "correct" search terms, either not enough or too many hits are returned (e.g. a Google search for "GSA" produces over one million results, mostly having nothing to do with geology). The underlying cause is a lack of "common sense knowledge" in current search tools. An emerging solution is through the "Semantic Web", an extension to the existing WWW environment, coordinated by W3C. The Semantic Web envisions common sense knowledge encoded directly into web pages via XML tags. These tags make reference to external ontologies that "define" terms and their mutual relationships. The Semantic Web for Earth and Environmental Terminology (SWEET) is a prototype funded by the NASA Earth Science Technology Office to explore such an application for the Earth sciences.

Ontologies are hierarchical, with child concepts containing properties that specialize their parent concept(s). The W3C has adopted the DAML+OIL ontology language as its standard Ontology Web Language (OWL). DAML+OIL is a specialization of RDF, itself a specialization of XML. DAML+OIL includes concepts from RDF (class, subclass, property, domain, range, etc.) supplemented with further semantic concepts (cardinality, inverse properties, synonyms, etc.). We selected DAML+OIL for this project.

Our ontologies include orthogonal concepts (space, time, Earth realms, physical quantities, etc.) supplemented with integrative science knowledge concepts (phenomena, events, etc.) Each orthogonal dimension constitutes a hierarchy of complexity (or richness); traversing down the associated tree follows the path of reductionism by adding additional details to more abstract concepts. An additional dimension “phenomena” is synergetic rather than orthogonal to the others. The phenomena entries describe synthesizing concepts that utilize elements from the other ontologies (e.g., earthquake has associated physical properties, Earth realms, etc.). Taken together, these complementary dimensions mirror the scientist's dual processes of reductionism and synthesis. This structure provides a semantic framework for classifying resources in terms of their underlying knowledge context.