Transcript uos--smith_20060314

Upper Ontology Summit
Tuesday March 14
The BFO perspective
Barry Smith
Department of Philosophy, University at Buffalo
National Center for Ontological Research
National Center for Biomedical Ontology
http://ontologist.com
Basic Formal Ontology
Basic Formal Ontology is a highest-common
denominator upper ontology designed to
support interoperability between domain
ontologies developed to support shared use
of scientific research data across disciplinary
boundaries
http://ontologist.com
Virtues of semantic interoperability
for scientific and clinical research
• allows data to be reused across disciplinary
boundaries
• allows communities to communicate across
disciplinary boundaries
• allows communities to reuse the same
software for managing their data
http://ontologist.com
What is ‘semantic’?
Model-theoretic (set-theoretic) semantics
provided on the basis of full formalization?
The meanings of terms are taken into account,
rather than just syntax?
The referents of terms are taken into account =
what terms refer to in reality, rather than in
some set-theoretic model?
http://ontologist.com
The State of the Art (in Biomedical
Ontology)
Many domain ontologies are pre-formal artifacts
Gene Ontology:
hemolysis =def. the causes of hemolysis
menopause part_of death
Such ontologies can be conveyed in a variety of
different formats; but much of their information
content is not formal in nature, or is not yet
formalized.
Upper ontologies can help bring advances where one
needs to work with pluralities of very large domain
ontologies, some with several million terms
http://ontologist.com
Virtues of upper ontologies for
scientific and clinical research
Early adoption of a high quality shared upper
ontology by different communities can mean
data is organized in compatible (‘semantically
interoperable’) ways
Formal definitions, axiomatization and formal
semantics of upper ontologies can add
reasoning power to be used both within and
between domain ontologies
Can import formal rigor into domain ontologies
(from the top)
http://ontologist.com
Upper ontologies require domain
ontologies with a need to
intercommunicate
Upper ontologies without domain ontologies
are inherently partial artifacts (like telephone
switchboards without subscribers)
http://ontologist.com
Upper ontologies require users
We agree that axioms and formal definitions are
an indispensable part of creating semantic
interoperability catalyzed through an upper
ontology
But an upper ontology will only be used by
domain scientists if we provide humanly
intelligible equivalents to these axioms and
formal definitions
These equivalent natural language axioms and
definitions should conform to the intuitions of
domain scientists
Upper ontology is something like the common
sense of domain scientists
http://ontologist.com
The methodology of annotations
Scientists perform experiments (including clinical trials)
yielding huge amounts of data
These experiments provide information about given
types of entities in reality, e.g. that genes and
proteins of this type are associated with this type of
clinical phenomena
Annotators create corresponding associations between
gene and protein names in databases and terms from
different domain ontologies such as the Gene
Ontology
Upper ontologies should help reasoning with this
annotation information
http://ontologist.com
Can we help?
Need for a registry of upper ontologies
National Center for Biomedical Ontology
(http://ncbo.us) Bioportal
= a registry of ontologies in the biomedical
domain, including different sorts of metadata,
supported by visualization and ontology
integration tools
http://ontologist.com
Can we help?
A registry of upper ontologies should contain evaluation
information, including information regarding actual
usage. (Cf. NCOR Evaluation Project)
There is a very small sub-lattice of concretely
expressed theories being used in actual integration
efforts.
Conclusion: this summit should contribute to greater
semantic interoperability (in significantly beneficial
ways) by showing how to reason across the upper
ontologies actually used (and thus to reason across
the different sorts of data annotated in their terms)
http://ontologist.com