As the proverb goes, preparation is key.
To plan for the upcoming scheduled technical discussion,
I'd like your input to focus the discussion down to a reasonable scope. (01)
Regards,
-- Nicolas. (02)
--------------------------------------------------------------------------------------------- (03)
Topic: Ontologies, reasoning & tagging (04)
Session moderator: Nicolas Rouquette (05)
Background: (06)
1) OWL is the de-facto common ground for relating external information
via tagging to an ontology where ontology-based reasoning has a useful
application-specific role.
2) Tool support for OWL-based development and reasoning has progressed
to the stage where OWL has become a practical technology, almost
"off-the-shelf"
3) There is plenty of lattitude for applying this technology in a
practical sense: this is both promising (e.g., innovative methodologies)
and concerning (e.g., project failures) (07)
Goal: assemble a panel of experts to discuss: (08)
a) selecting what useful functions can ontology-based reasoning have
w.r.t. application-specific issues, questions, problems (09)
examples: (010)
a.1) ontology development support
=> reasoning: -- see classification here:
http://protege.stanford.edu/conference/2004/schedule2.html#thursday
=> tools -- see a list here: http://www.w3.org/2004/OWL/#paa (011)
a.2) semantic query
=> reports, extracting a specific subset of an ontology -- see document
generation here:
http://protege.stanford.edu/conference/2004/schedule2.html#thursday
=> tools -- see list above
=> query languages & tools
comparison: http://www.aifb.uni-karlsruhe.de/WBS/pha/rdf-query/
the new kid on the block: sparql
see: http://www.aifb.uni-karlsruhe.de/WBS/pha/rdf-query/
why should you care? -- see:
http://lists.w3.org/Archives/Public/www-rdf-rules/2005Apr/0001.html (012)
a.3) validation / verification
=> consistency/integrity checking
=> this is big problem but one that can quickly escalate into fierce
arguments --- see:
http://lists.w3.org/Archives/Public/www-ws-desc/2005Mar/0026.html
e.g., is there a sensible strategy to use a (meta) ontology with which
one can describe the kind of consistency/validation/... criteria we want
a (lower) ontology to have? (013)
b) handling expressiveness / modeling / representability issues (014)
=> sometimes, OWL isn't enough (because of limited expressiveness) to
describe complex important application-specific notions.
eg. -- see the session on "biomedical ontologies" here:
http://protege.stanford.edu/conference/2004/schedule2.html#thursday (015)
examples: (016)
b.1) a possible next step: use an "upper" ontology that is sufficiently
expressive to adequately capture such complex application-specific notions. (017)
=> how? (018)
b.2) annotation strategy
concern: annotations are currently "ignored" because the meaning of an
annotation depends on what annotations are used for and what meaning is
assigned to them
possible strategy: develop an "ontology" that describes the meaning of
an annotation
motivation1: the meaning of an annotation is clear
motivation2: use the meaning of an annotation to select which
annotation(s) are relevant/useful for a specific reasoning purpose (019)
problems:
=> including an annotation ontology can easily lead to a logical theory
that is no longer OWL-DL but OWL-Full (020)
c) selecting among strategies for capturing application-specific
semantics into a useful form for such reasoning functions: (021)
c.1) tagging application-specific domain concepts & data with
ontological annotations
c.2) splitting the application-specific domain into different ontologies
(e.g., separating the simple stuff from the complex stuff)
c.3) formalizing the complex stuff with a different kind of theory (022)
=> annotation is a common example of a practical strategy (see best
practices guidelines for representing classes as property values) (023)
=> is using an "upper" ontology like SUMO, DOLCE or PSL for annotation
purposes a sensible strategy? (024)
possible motivations: (025)
* reusing/leveraging high-quality ontologies whose semantics have been
carefully and rigorously defined.
* the reasoning problem involves domain-specific knowledge beyond the
scope of the ontology
(e.g., planning as a form of application-specific reasoning style in
the context of web services (026)
http://www.daml.org/services/use-cases/language/swsl-usecase/Composition.htm) (027)
=> the semantics of such "upper" ontologies (e.g. SUMO, DOLCE, PSL)
require languages that are more expressive than OWL (e.g., KIF)
catch 22: reasoning in the context of expressive ontologies can
quickly turn into an excercise in theorem prooving
theorem prooving isn't practical in many cases: theorem proovers
require a lot of "guidance" from a sufficiently skilled / knowledgeable user (028)
problems:
=> are there circumstances in which theorem prooving can be reasonably
automated w.r.t. ontologies that are formalized in a suitable way? (029)
alternative strategies without theorem prooving: (030)
=> if reasoning is the means to an end (e.g., answer a question about an
ontology, a model built w/ the ontology, ...)
can we extract the relevant portion of the ontology that is
necessary w.r.t. reasoning about a specific question? (031)
possible motivation: reasoning w/ the full ontology may be overkill;
reasoning w/ the relevant portion of the ontology ought be a simpler and
tractable problem (032)
questions:
=> are there practical strategies for "extracting" a subset of an
ontology that is logically relevant & useful w.r.t. a specific reasoning
problem? (033)
--------------------------------------------------------------------------------------------- (034)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Unsubscribe/Config:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (035)
|