Overview
During the last decade the Semantic Web community has established
basic standards (e.g., RDF, OWL) for representing data and ontologies.
However, these formalisms have been designed for computer processing
and data exchange rather than for easy comprehension by people;
OWL (the standard ontology language) in particular presents many
conceptual hurdles for entry-level users. Wide access to the Semantic
Web therefore depends on the development of transparent editing
and viewing tools.
The main requirement for transparency is use of a medium that
is accessible to a wide non-specialist audience; for this purpose,
the obvious choice is natural language (e.g., English, French,
Italian) text, supported where appropriate by familiar quasi-linguistic
media like tables. We seek a set of principles by which logic-based
descriptions (especially ontologies on the Semantic Web) can be
associated with transparent formulations in natural languages.
By applying these principles, we will demonstrate proof-of-concept
tools that allow subject-matter experts to edit and query metadata
on the Semantic Web through a reliable natural language interface.
SWAT innovates by using the Open University's generation-based
technology which obviates any need for parsing or interpretation
of text, building on existing work (coordinated by an OWL 2 task
force) to use controlled natural language input for knowledge
formation on the Semantic Web.
Our techniques
Our techniques and tools are expected to be of interest to Semantic
Web users who wish to encode or query knowledge without deep understanding
of RDF, OWL, or the other specialist languages. We will focus
on some typical use cases in the areas of health provision (NHS,
Siemens, and WHO) and e-Science (MyGrid), and evaluate our results
in a new application in the travel domain which will be used in
Manchester University's OWL tutorial course.
Results will be made available to the wider Semantic Web community
(e.g., through the OWL Controlled Natural Language task force
coordinated by Ordnance Survey, and Stanford University's Protege
site).
|
