Web Ontology Segmentation Analysis, Classification and Use Julian - - PowerPoint PPT Presentation

▶
web ontology segmentation
SMART_READER_LITE
LIVE PREVIEW

Web Ontology Segmentation Analysis, Classification and Use Julian - - PowerPoint PPT Presentation

Jan 01, 2023 103 likes •412 views

Web Ontology Segmentation Analysis, Classification and Use Julian Seidenberg Alan Rector jms@cs.manchester.ac.uk rector@cs.manchester.ac.uk Ontology ?! Definition: An ontology describes concepts in a domain of interest and the

slide-1
SLIDE 1

Web Ontology Segmentation

Analysis, Classification and Use Julian Seidenberg jms@cs.manchester.ac.uk Alan Rector rector@cs.manchester.ac.uk

slide-2
SLIDE 2

Ontology ?!

Definition: An ontology describes concepts in a domain of interest and the relationships that hold between them.

2

slide-3
SLIDE 3

Use

Schema mapping Knowledge capture Information inference Intelligent user interfaces and ...

3

slide-4
SLIDE 4

4

global interoperability

slide-5
SLIDE 5

5

large domain ontologies

Social Science Medicine Biology Astronomy Ecology Particle Physics

slide-6
SLIDE 6

6

number of concepts time to classify Gene Ontology 20,000 220 sec. GALEN 23,000 stack overflow! NCI-Thesaurus 42,000 815 sec. FMA 70,000 stack overflow! SNOMED-CT 364,000 stack overflow!

(reasoner: Racer 1.7)

slide-7
SLIDE 7

4000 years ago ...

slide-8
SLIDE 8
slide-9
SLIDE 9
slide-10
SLIDE 10
slide-11
SLIDE 11

3 styles of segmentation

slide-12
SLIDE 12

partitioning

12

slide-13
SLIDE 13

querying

? ?

13

slide-14
SLIDE 14

?

traversal

14

slide-15
SLIDE 15

Background: superclasses as links

15

Heart ⊆

InternalOrgan ∃ isPartOf . CardiovascularSystem

GeneralisedStructure CardiovascularSystem BodySystem AbstractStructure isA isA isA Heart InternalOrgan BodyStructure isA isA isA

∃ isPartOf

slide-16
SLIDE 16

Segmentation by Traversal

16

Target of Segmentation cross-link cross-link Target of Segmentation cross-link cross-link Target of Segmentation cross-link cross-link Target of Segmentation cross-link cross-link Target of Segmentation cross-link cross-link

slide-17
SLIDE 17

Test Case

17

  • Big

23,000 classes

  • Complex

10,000 definitions 30,000 relationships

  • Representative

basis of many future systems

slide-18
SLIDE 18

segment of the ‘Heart’

18

5,000 10,000 15,000 20,000 25,000 number of classes Original Segment

slide-19
SLIDE 19

19

Boundary Limiting

slide-20
SLIDE 20

Limitation

(Boundary Limiting)

20

missing information ↓ lost inferences

slide-21
SLIDE 21

boundary limit vs. size

21

1,500 3,000 4,500 6,000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 30

number of classes

boundary depth

Good control over size But does not

  • ptimize

classification

slide-22
SLIDE 22

Target of Segmentation hasPart hasPart hasLocation hasLocation

Property Filtering

22

Target of Segmentation hasPart hasPart hasColor hasStatus

(partitive filter)

slide-23
SLIDE 23

Limitation

(Property Filtering)

23

modified definitions ↓ unexpected inferences

slide-24
SLIDE 24

Definition Transformation

24

SkinOfFrontalScalp ≡ (SkinOfScalp ∩ ∃ hasProximity . FrontalBone) SkinOfFrontalScalp ≡ SkinOfScalp SkinOfFrontalScalp ⊆ SkinOfScalp

slide-25
SLIDE 25

number of class

25

1,500 3,000 4,500 6,000 b a s i c c

  • n

s t r u c t i v e p a r t + f u n c + m

  • d

l

  • c

a t i v e p a r t i t i v e s t r u c t u r a l f u n c t i

  • n

a l m

  • d

i f i e r

slide-26
SLIDE 26

classification time

26

  • 3

19 41 63 85 b a s i c c

  • n

s t r u c t i v e p a r t + f u n c + m

  • d

l

  • c

a t i v e p a r t i t i v e s t r u c t u r a l f u n c t i

  • n

a l m

  • d

i f i e r seconds

slide-27
SLIDE 27

seconds to create segment

27

Modifier Functional Structural Partitive Locative Part + Func + Mod Constructive 15 30 45 60 75 90 105 120

loading scanning extraction saving

slide-28
SLIDE 28
  • Segmentation by traversal

Semi-automatic Works well with densely interconnected ontologies Independently coherent subset (not simply a view)

  • Boundary limiting

Accurately controls size

Summary

  • 3 different styles of segmentation

Partitioning, querying, & traversal

28

  • Property filtering

Significantly increases tractability Facilitates ontology profiling

slide-29
SLIDE 29

Questions ?

29

Julian Seidenberg jms@cs.manchester.ac.uk