Why I need both OWL/DLs

1
Why I need both OWL/DLs Frames

• Alan Rector
• Medical Informatics GroupBio Health Informatics
  ForumDepartment of Computer ScienceUniversity
  of Manchesterrector_at_cs.man.ac.uk
• oiled.man.ac.uk
• www.bhif.man.ac.ukwww.mig.man.ac.uk

2
CO-ODE/HyOntUseBringing Protégé and OWL/OilEd
Together
OilEdThe de factostandard editor
forDAMLOIL/OWL/logic-based ontologies
Plus methodsfromOpenGALENPENPAD AKT
Protégé The de facto standard environmentfor
frames
Not as easy as it looks!
3
Very Brief History of OWL

• OIL European approach Description Logics in
  Frame Clothing
• Initial OilEd - Manchester
• DAML DARPA Agent Markup Language
• DARPA
• DAMLOIL
• First joined up approach- EUDARPA
• OWL
• Emerging W3C WebOnt Standard
• 3 Flavours Lite, DL, and Full still
  evolving
• I work mostly with the subset of DL that works
  with existing classifiers
• De facto standard way to apply logic-based
  ontologies
• OilEd still the main editor but new efforts e.g.
  PROTÉGÉ-OilEd/OWL tab coming

4
Why I need both OWL/DLs and Frames

• Build real large-scale knowledge intensive
  applicationsOntology Anchored Knowledge Bases
• Fractal Adaptation
• Rebuild PENPAD introduction
• GRAIL is essentially a hybrid Frame/DL system
• Build robust auditable applications
• Get the ontology right
• Meta data and provenance
• Achieve sufficient abstraction for re-use
• From application ontologies to domain ontologies
• Get the right answer to the intended question
• Do I mean Is it possible or Is it true?
• Do only what is needed

5
Why I need both OWL/DLs Frames

• To Build Knowledge Intensive applications
• Knowledge bases anchored on ontologies supporting
  information resources
• Meta data with everything

6
Why I need OWL/DLs

• Maintain large, complex ontologies/terminologies
• Parsimonious ontologies - Conceptual lego
• Avoid combinatorial explosions
• Strong semantics for Reasoning about Subsumption
  Normalisation
• Modularity
• Avoid inheritance conflicts (Nixon Diamonds)
• but it lacks
• Meta data
• Defaults exceptions
• Reflective queries
• Reasoning/Querying with individuals
• Other forms of reasoning arithmetic,
  coordinate/unit transformation,
• and it does too much
• Complete reasoning about what is possible when I
  need predictable reasoning about what is true
• Domain range checks

7
Why I need Frames/Protégé

• Manage Metadata, Contingent knowledge
  Individuals
• Knowledge about Knowledge
• Defaults exceptions classic frame reasoning
• Individuals
• Reflective queries ask about the knowledge base
  itself
• Hybrid reasoning
• Easy to integrate special purpose solutions for
  special purpose problems
• Easy to extend expressiveness for queries
• but it lacks
• Parsimonious representation No Lego
• Strong semantics for subsumption
• Reasoning about what is possible rather than just
  what is

8
I need to experiment with much more
metadataProvenance, provenance, provenance
9
Maintaining large Ontologies Conceptual Lego
SNPolymorphism of CFTRGene causing Defect in
MembraneTransport of ChlorideIon causing Increase
in Viscosity of Mucus in CysticFibrosis
Hand which isanatomicallynormal
OpenGALEN OWL
10
Whats in a Logic based ontology?

• Primitive concepts - in a hierarchy
• Described but not defined
• Properties - relations between concepts
• Also in a hierarchy
• Descriptors - property-concept pairs

• qualified by some, only, at least, at
  most
• Defined concepts
• Made from primitive concepts and descriptors
• Axioms
• disjointness, further description of defined
  concepts
• A Reasoner
• to organise it for you

• Meta data
• Contingent Knowledge
• Defaults Exceptions
• Reflective queries
• Individuals
• Hybrid reasoning

OWL / DLs
Frames
11
OWL/Logic Based Ontologies The basics
Primitives
Descriptions
Definitions
Reasoning
Validating
Thing
red partOf Heart
red partOf Heart
(feature pathological)
12
The Key Normalising (untangling) Ontologies
13
The Key Normalising (untangling)
OntologiesMaking each meaning explicit and
separate
PhysSubstance Protein ProteinHormone
Insulin Enzyme Steroid
SteroidHormone Hormone ProteinHormone
Insulin SteroidHormone
Catalyst Enzyme
PhysSubstance Protein ProteinHormone
Insulin Enzyme Steroid
SteroidHormone Hormone
ProteinHormone Insulin
SteroidHormone Catalyst Enzyme
...and helping keep argument rational and
meetings short!
Hormone Substance playsRole-HormoneRole Pro
teinHormone Protein playsRole-HormoneRoleS
teroidHormone Steroid playsRole-HormoneRole
Catalyst Substance playsRole
CatalystRole Insulin ? playsRole HormoneRole
Enzyme ?? Protein playsRole-CatalystRole
14
The benefitsAvoiding combinatorial explosions

• The Exploding Bicycle From phrase book to
  dictionary grammar
• 1980 - ICD-9 (E826) 8
• 1990 - READ-2 (T30..) 81
• 1995 - READ-3 87
• 1996 - ICD-10 (V10-19 Australian) 587
• V31.22 Occupant of three-wheeled motor vehicle
  injured in collision with pedal cycle, person on
  outside of vehicle, nontraffic accident, while
  working for income
• and meanwhile elsewhere in ICD-10
• W65.40 Drowning and submersion while in bath-tub,
  street and highway, while engaged in sports
  activity
• X35.44 Victim of volcanic eruption, street and
  highway, while resting, sleeping, eating or
  engaging in other vital activities

15
The benefitsModularisationBridging Scales and
context with Ontologies
Species
Genes
Function
Disease
16
Benefits Fractal Indexing on multiple axes

• Indefinite customisation from a finite knowledge
  base
• Consistent application of policies
• Fail soft always produce something plausible
• Multiple axes of specialisationNormalised
  ontologies produce few inheritance conflicts
• Condition
• Use case
• Task
• User type
• Setting
• Medium
• Browser, PDA, WAP, Thick client,

17
Example Fractal tailoring of Forms/Guidelines/Pro
cedures

• Cough
• Initial evaluation in general practice
• routine evaluation in general practice
• routine evaluation by nurse in general practice
• Home monitoring
• Cough in patient with TB
• as above
• In chest clinic
• In Dr Jones chest clinic
• In Dr Jones chest clinic seen by a trainee
• 100 diseases x 10 complications x 5 settings x 5
  user types x 5 tasks ? 25000 situations
• Do you really want to enumerate them by hand?
  maintain them?

18
PENPADFractal Tailoring of fail soft forms
19
(No Transcript)
20
Fractal tailoring forms for clinical trials
Hypertension
Hypertension
Idiopathic Hypertension
Idiopathic Hypertension
In our companys studies
In our companys studies
In Phase 2 studies
In Phase 2 studies
21
Other Fractal Indexing Tasks

• Mapping to between coding systems and ontologies
• From logical to alogical systems e.g. ICD10
• All ICD excludes come automatically
• Drug interactions and contraindications and usage
• Contingent knowledge not part of necessary
  nature of drug
• Help systems
• Gather all relevant information from all levels
• Selecting relevant guidelines and trial protocols

22
But it is not trivial

• OWL and Frame paradigms are more different than
  they look
• OWL is concerned with axioms
• Protégé is concerned with facts
• Structure of graph
• OWL focuses on restrictions
• Roughly the allowed classes/existent classes
  facets
• Class-instance distinction principled
• Protégé focuses on values
• Meaning of a class value ambiguous used
  differently in different applications
• Class instance distinction application dependent
• OWL supports ONLY an ontology one kind of
  reasoning
• Protégé supports knowledge bases potentially
  many kinds of querying
• but not OWLs open world reasoning!

23
Classifying and QueryingOnly doing the reasoning
necessary

• Classifying OWL, DLs,
• What must be true or false
• In any extension of this world consistent with
  axioms
• related to modal logics
• Negation impossibility (unsatisfiability)
• Open World
• Computationally expensive
• Limits expressivity
• Persistent
• Querying PAL, Query tab, SQL,
• What is true or false
• In this world about which we know facts
• Negation failure
• Closed World
• Computationally relatively cheap (usually)
• Ephemeral

24
Classifying and QueryingThe Pizza Example

• MyPizza Pizza hasTopping Peppers
  hasTopping Mushrooms
• Is MyPizza a vegetarian pizza?
• Classification/OWL
• No not necessarily, you havent said it
  doesnt have meat
• Negation as impossibility
• open world
• Querying/Database
• Yes I cant find any meat
• Negation as failure
• closed world

25
User Oriented Ontology DevelopmentCO-ODE
HyOntUse

• New projects under the UK JISC/EPSRC joint
  initiative on Semantic Web Autonomic Computing
  Initiative parallel with US NLM/NCI funding
• Collaboration - Manchester, Stanford,
  Southampton/Epistemics
• Integrate Bridge the Gaps
• Frames Metaknowledge Protégé
• Plug Play environment
• Visualisation, DAGs, Constraints,
• Logic based domain ontologies
  DAMLOIL/OWL/OilEd
• User oriented debugging and visualisation
• Views Perspectives GALEN
• User oriented design / Knowledge Elicitation
  AKT/Southampton

26
CO-ODE/HyOntUse Consortium

• Manchester CS
• With thanks to Sean Bechhofer, Carole Goble,
  http//oiled.man.ac.uk
• Stanford Medical Informatics
• With thanks to Holger Knublauch, Ray Fergerson,
• Southampton Advanced Knowledge Technologies
• With thanks to Nigel Shadbolt Clive Embury
  (Epistemics)
• and all of you
• Help to improve usability, visualisation,
  applications, ,
• Help us Help yourselves Join in Invite
  others!

PS Post Doc Needed!