No Slide Title

Download Report

Transcript No Slide Title

DAML+OIL is not Enough
Sean Bechhofer, Ian Horrocks, Carole Goble
Information Management Group
University of Manchester, UK
http://img.cs.man.ac.uk
Semantic Web Working Symposium
Stanford, USA
July 30 - August 1, 2001
Take Home Message
• OIL was good and popular.
• DAML+OIL is still good, but….
• The move away from a frame-like representation may
make modelling harder.
• Even more need for good tool support.
OIL: The Three Roots
• Frame-based Representations;
• Description Logics;
• Web Languages.
Description Logics:
Formal Semantics &
Reasoning Support
Frame-based Systems:
Epistemological Modelling
Primitives
OIL
Web Languages:
Ontologies &
XML- and RDF-based
Lexicons
syntax
OIL: Frames
• Modelling primitives based on classes (frames) with
local attributes.
• Provide a natural and friendly representation for
modellers. Has proved popular with, e.g.
bioinformaticians.
 Sometimes a lack of well-defined semantics.
UK-Animal-lover
Superclass
Person
Has-pet
> 3 Animal
Lives-in
UK
OIL: Description Logics
• Describe knowledge in terms of concepts and roles.
• Generally supply a range of concept forming operators:
– boolean connectives and, or, not and role restrictions
• Well-defined, formal semantics allowing the use of
reasoners.
 Hard(er) to interact with directly.
UK-Animal-Lover 
Person 
.livesin UK 
atleast 3 has-pet Animal
OIL: Web languages
• A representation language needs a delivery format and
concrete syntax.
• OIL originally specified using XML DTD, later with RDF
Schema.
• Ensures compatability with existing tools and gives
RDFS-aware applications access to OIL ontologies.
 RDF isn’t really human-readable,
thus tools are needed.
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:dc="http://purl.org/dc/elements/1.1/"
xmlns:oil="http://www.ontoknowledge.org/oil/rdf-schema/2000/11/10-oil-standard#"
xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#">
<rdf:Description rdf:about="" dc:creator="Sean Bechhofer" dc:description="An Ontology about animals, built
with OilEd"
dc:description.release="1.0" dc:identifier="id" dc:language="OIL" dc:title="Animals"
dc:type="ontology" />
<rdfs:Class rdf:about=“http://cohse.semanticweb.org/ontologies/elephants.rdfs#elephant">
<rdfs:comment>Big grey animals with a trunk and large ears. Belonging to the family of
pachyderms,
elephants can be subdivided into the class of Indian elephants, i.e. those coming from India, and
African elephants, those coming from Africa. One of them has bigger ears than the other, but I can
never remember which one it is.</rdfs:comment>
<oil:hasPropertyRestriction rdf:resource="_anon12" />
<oil:hasPropertyRestriction rdf:resource="_anon13" />
<rdfs:subClassOf rdf:resource="file:F:/Projects/OIL/OilEd/ontologies/elephants.rdfs#animal" />
</rdfs:Class>
</rdf:RDF>
DAML+OIL
•
•
•
•
Bringing together DAML’s ontology language and OIL.
Keeps OIL’s mapping to Description Logic.
Keeps the Web language mapping.
Moves away from frame-style language.
– General axioms rather than frame descriptions.
– Essentially an alternative syntax for a DL.
• OilEd: a simple ontology editor.
• Originally developed to demonstrate the use of a
reasoner with OIL.
• Exploits the frame-based
paradigm of OIL. But also
extends pure frames,
allowing arbitrary expressions.
• Download from:
http://img.cs.man.ac.uk/oil
• Lunchtime demo tomorrow!
DAML+OIL vs. OIL
• The semantics is still preserved, but..
 It becomes harder to reconcile frame-based
descriptions with their underlying representation in
DAML+OIL.
 Information about how the descriptions have been
structured is lost.
DAML+OIL applications
What does it mean for an application to use DAML+OIL?
1. writing out a model as DAML+OIL;
2. reading it back in from the DAML+OIL;
3. reading someone else’s DAML+OIL.
White Van Man: frames
class-def defined white_van_man
subclass-of man
slot-constraint drives has-value
(van and
(slot-constraint has_colour has-value
(one-of white)))
implies white_van_man aggressive_driver
Basic Definition
Axioms
White Van Man: axioms I
equivalent white_van_man
(man and
slot_constraint drives has-value
(van and
(slot-constraint has_colour has-value
(one-of white)))
implies white_van_man aggressive_driver
• How do I know which axiom
corresponds to the
definition?
Axioms
White Van Man: axioms II
implies white_van_man man
implies white_van_man
slot_constraint drives has-value
(van and
(slot-constraint has_colour has-value (one-of white)))
implies (man and
slot_constraint drives has-value
(van and (slot-constraint has_colour has-value
(one-of white)))
white_van_man
implies white_van_man aggressive_driver
Axioms
Is this a problem?
• Not for delivery and use of ontologies.
• Modelling and exchange require a common understanding of the
model.
• If we adopt the frame-based paradigm for modelling, we may not
be able to consistently recreate the original model from the
axioms.
• Tools may have to guess the modeller’s intended structure.
• Good understanding cannot be ensured by meaning preservation
[Euzenat]
Solutions?
• Extend the schema locally (i.e. tool-specific information)
to represent the structure of the information.
• Fine if everyone uses the same tool.
 Moves away from the standard, and introduces bespoke
representations that may not be easily exchanged.
Concluding Remarks
• Some of OIL’s good features may have been lost in
DAML+OIL
• RDF syntax provides freedom of expression, but…
– Provides no clear structure
– May make modelling harder
– Is notoriously baroque
• Requirement for tools, but…
– Lack of structure has serious implications for tool builders