Hi All,
I recently attended a conference on schemas and ontologies. Here is a
summary of the main points.
Ed
Schemas and Ontologies: Building a Semantic Infrastructure for the GRID
and Digital Libraries
Date 16 May 2003-05-22, Venue e-Science Institute, 15 South College Street.
The Schemas and Ontologies seminar was organised by UKOLN and the
e-science Core Programme. The aim of the event was to help bring GRID
and digital library implementers together to consider approaches to
developing, expressing and sharing schemas and ontologies.
The event consisted of speakers explaining some theoretical aspects of a
semantic infrastructure and practical examples of ontology and schema
use in medicine, biology and educational metadata. There were also
discussion breakout groups looking at: Barriers to sharing ontologies;
Software tools and shared services; The process of building a
community-led ontology.
Formally speaking an Ontology “defines terms used to describe and
represent an area of knowledge” (Ref 1). One way of thinking about an
ontology is to consider it as: a collection of objects that are linked
together in certain ways. This means that the word ontology can define a
range of structures from simple taxonomies to metadata schemas, to
logical theories.
A very good introduction to ontologies can be found at (Ref 2). This
uses a wine ontology as an example
Ontologies can be described using web ontology languages. The most
recent one is OWL (Ref 8) and it has been formed from DAML and OIL.
A Practical example of ontology use (in the field of medicine) was given
by Dr Jeremy Rogers from the University of Manchester He explained how
in the past medical conditions had been described by long descriptions.
Current systems use combinations of words to form concepts. So one might
form a concept by selecting several words from several lists. This sort
of information is easier for machines to read than lengthy descriptions.
The cost of building the terms and rules is large and there can be
problems with interpretation of results.
Professor Carole Goble, University of Manchester, talked about the use
of ontologies for carrying semantics in the context of The Grid with
reference to projects myGrid(Ref 3) and Geodise (Ref 4). Issues
discussed included looking at tools and services for supporting
biological sciences and ontologies as a means of integrating and
exchanging resources.
It is also important to make sure communities are aware of other
practices in order to facilitate resource sharing and reuse. Registries,
which store details of schemas and ontologies, can encourage sharing and
reuse.
Rachel Heery of UKOLN, talked about the need to use registries and
promote publishing and sharing schemas. The idea is to manage schemas in
such a way that that reuse and sharing of metadata is encouraged but to
allow for extensibility and meet local requirements. This was
supplemented by a practical demonstration of the MEG Registry by Pete
Johnston of UKOLN. This Registry allows users to register educational
metadata schemas and elements and make use of existing schemas, elements
and vocabularies. For further details see (Ref 7)
Doug Tudhope of the University of Glamorgan gave a talk on Knowledge
Organisation Systems (KOS). This covered a review of current Digital
Library work on KOS and research into connections between ontologies and
the semantic web. The two main sources of information for the talk were
NKOS: Networked Knowledge Organisation Systems/Services (Ref 5)and
SEMKOS FP6 IP Proposal (Ref 6)
NKOS is devoted to the discussion of the functional and data model for
enabling knowledge organization systems (KOS), such as classification
systems, thesauri, gazetteers, and ontologies, as networked interactive
information services to support the description and retrieval of diverse
information resources through the Internet. SEMKOS is involved in
research into connections between KOS and The Semantic Web.
DISCUSSION GROUPS
The discussion groups, which were set up, covered many of the issues
that are important to establishing and maintaining ontologies. It is
interesting to note that almost all the points mentioned apply to
metadata schema implementation as well.
Thee are a lot of barriers to sharing ontologies:
· Cost of creation and maintenance is very high. The medical ontology
discussed by Jeremy Rogers contains 20,000 concepts and took 15
work-years to create.
· The technology is immature and there are not many tools available.
· People need to be aware of the costs and benefits about using and
sharing ontologies.
· It is difficult to ensure quality control.
Several suggestions about how to encourage uptake were given
· Ontologies should be free if used for non-profit purposes
· Further tools are necessary to support ontology building and browsing
· Clear applications and demonstrations need to be given to promote use.
· Establishing ontolgies requires a lot of money, time and effort.
Professor Goble and Dr Rogers gave examples of successful ontologies and
how they were set up. For the medical application, the project was well
funded for set up and maintenance. For the genetics ontology, it started
off with a small group of enthusiasts that gradually became more
established and received funding.
· It is important to be systematic when creating an ontology. Adding new
terms can either be done by group decision or by having one person in
charge who decides what should go in (referred to as the “benign
dictator” method). Both methods have benefits and problems.
· It is important to make sure that end users will be able to make
practical use of the ontologies and that user interfaces are not too
complicated.
REFERENCES
1) Quote from Jeff Heflin, Web Ontology Language Use Cases and
Requirements, W3C, February 2003, http://www.w3.org/TR/webont-reg/
2)
http://www.ksl.stanford.edu/people/dlm/papers/ontology-tutorial-noy-mcguinness-abstract.html.
3) http:/www.mygrid.org.uk
4) http://www.geodise.org
5) http://nkos.slis.kent.edu/
6) http://www.lub.lu.se/SEMKOS/
7) http://www.ukoln.ac.uk/metadata/education/regproj/.
8) http://www.w3.org/TR/2002/WD-owl-guide-20021104/
|