5. Ontological Engineering

5. Ontological Engineering .

5.1. Ontology: new old conception 5.2. Ontological engineering 5.3. Ontology formalization 5.4. Ontologies on the Web Course on KE, part 5, by Gavrilova T.

1

5.1. Ontology: new old conception In philosophy • Ontology - the branch of philosophy that studies the nature of existence, opposite to

• Epistemology – the branch of philosophy that

investigates the origin, nature, methods, and limits of human knowledge.

Course on KE, part 5, by Gavrilova T.

2

Ontologies Now ontology is not only a philosophical discipline studying the being. In computer science ontology defines the basic terms and relations comprising the structured vocabulary of a topic area. “Ontology is an explicit specification of a conceptualisation or a hierarchically structured set of terms for describing a domain that can be used as a skeletal foundation for a knowledge base”. (Gruber, 1993) Course on KE, part 5, by Gavrilova T.

3

Other definitions • Ontology as an informal conceptual system. • Ontology as a formal semantic account. • Ontology as a representation of a conceptual system via a logical theory. • Ontology as the vocabulary used by a logical theory. Course on KE, part 5, by Gavrilova T.

4

Visual approach Step 1: Create and name concepts A E

G

F D

B

J H

I C

K

Course on KE, part 5, by Gavrilova T.

5

Visual Approach Step2: Link concepts A B

C

D

I

J

F H

G E Course on KE, part 5, by Gavrilova T.

K

6

Visual Approach Step 3: Ladder concepts A F H

E G

C B

I

J

K D

Course on KE, part 5, by Gavrilova T.

7

Ontology as Hierarchical Conceptual Structure (WHAT -knowledge) A

C

S

A

F

W

D

N

B

D H

L G

W

R

S

U

C

G

M

E

M N

R

U

H

L

Course on KE, part 5, by Gavrilova T.

F

B

E 8

Ontologies System SubSystem1 Unit 1 Object 1 Value1 Value2

SubSystem2 Unit 2

Object 2

Object 1

SubSystemN Unit N

Object 2

Course on KE, part 5, by Gavrilova T.

Object N 9

Why we need Ontologies? O b je c tiv e o f O n to lo g y T e a c h in g

U n d e rs ta n d in g fo r W o rk in g

M anagem ent o r C o n tro l

S e a rc h , A c c e s s a n d N a v ig a tio n

Course on KE, part 5, by Gavrilova T.

R e s e a rc h

10

Ontologies Relationship type

Owner

Personal

Group Corporate

Classobject

Representation of

Has-Part

Attribute/value

Association

Technologies

Derivation

Customers

Target or Objective

Organization Services

Businessprocesses Course on KE, part 5, by Gavrilova T.

11

How to present ontology Ontology mapping Ontology as

Traditional hierarchy

List of properties

Contents (catalogue)

Relational table

Course on KE, part 5, by Gavrilova T.

Hypertext structure

12

Subjectivity of ontology “SOFTWARE- ontology N1”

Course on KE, part 5, by Gavrilova T.

13

Ontology ”Software” N2

Course on KE, part 5, by Gavrilova T.

14

Ontology as content structure

Course on KE, part 5, by Gavrilova T.

15

“Presentation” – 1-level-ontology

Course on KE, part 5, by Gavrilova T.

16

“Presentation” -2-level ontology

Course on KE, part 5, by Gavrilova T.

17

Mind map or Ontology?

Course on KE, part 5, by Gavrilova T.

18

Mind map STUDENT N2

Course on KE, part 5, by Gavrilova T.

19

Test N 29 • “Student” ontology design

Course on KE, part 5, by Gavrilova T.

20

5.2. Ontology engineering From Mind map to Ontology

• Ontology is a well-structured holistic homogeneous mind map • Ontology is a laddered semantic net • Ontology is a bridge mind tool from mental net to more scientific and structured hierarchical presentation form • Course on KE, part 5, by Gavrilova T.

21

Foundations of Ontology Engineering Three Modeling Levels Elicitation

Goals and Requirements Level (WHY) Conceptual Model Level (WHAT)

Structuring Formalization

Specification Level (HOW) Course on KE, part 5, by Gavrilova T.

22

Ontologies

Course on KE, part 5, by Gavrilova T.

23

Main Relations ( partly from work of Asuncion Gomez- Perez asun@ fi. upm. es) • Between classes: Subclass (superclass)- of Subclass- partition • between objects (concepts) and classes Instance- of (AKO -A-kind-of) Has- Instance • between objects has part has attribute Course on KE, part 5, by Gavrilova T.

24

Main types of ontologies Name • Taxonomy (place in class) • Partonomy (consistence) • Attributive frame (main features) • Derivative Tree (history and family) • Associative Map (close associations or mind maps) • … • Mix

Relationship AKO Has-part Attribute (value) Parent-son (genealogy) Association

variety

Course on KE, part 5, by Gavrilova T.

25

Ontology mind map

Course on KE, part 5, by Gavrilova T.

26

Partonomy (“Has-part”) ontology

Course on KE, part 5, by Gavrilova T.

27

AI ontology AI

Creativity

Games

Theoretical & Formal Foundations

Music, poetry

Natural Language Processing

Text Analysis

Machine Learning

Speech Recognition

Connectionism

Knowledgebased systems

Robotics

Symbolic learning

Expert systems

K. Acquisition

Data Mining

Applications

AI-software

Languages

Tools

Knowledge Engineering

Genetic algorithms

Neural Network

Distributed AI

K.Structuring

R.representation

Concept Mapping Knowledge Discovery

Knowledge Elicitation

Course on KE, part 5, by Gavrilova T.

Ontological engineering

28

TAXONOMY example

Tatiana Gavrilova

St St--Petersburg Petersburg State State Technical Technical University Universit y

Knowledge Elicitation Techniques Taxonomy KE KE Techniques Techniques

Comm unicativ unicativ ee Groupe - Round Round Table Table - Brain Brain Storm Storm - Games Games

Textological Textological - Document Document Analysis Analysis - Literature Literature Analysis Analysis

Indiv Indiv idual idual Passiv e -- Ob servations -- Verbal Reports Reports -- Lecture s

Activ Activ e - Interview Interview - Quetionnaire Quetionnaire - Role Role Games Games

Course on KE, part 5, by Gavrilova T.

29

Attributive ontology Tree

Geobotanical features

Building features

Physical features

Appearance

Height Density

Humidityproofness

Leaf features

Colour Form

Course on KE, part 5, by Gavrilova T.

30

Derivative ontology (Genealogy) 8086 8088 80186 80286 80386 80486 80586 i ntel Pentium I Pentium MX

68000 motorola

Pentium XEON

Pentium II

68010 68030

Pentium III Course on KE, part 5, by Gavrilova T.

31

Michelangelo

Rafael Perujino

Signorelli

Donatello

Fra Angelico Boticelli Titian

Veroccio

Girlandaio Tintoretto

Mantegna

Jac.Bellini Giorgione P.Venetiano

Masaccio

Veronese

Carpaccio

Jiov.Bellini

Giotto Leonardo

Course Pinturiccio on KE, part 5, by Gavrilova T.

Cimabue

32

Cimabue

XIII

Giotto

P. Venetiano

Masaccio

Jac.Bellini

Donatello

Fra Angelico

Mantegna Veroccio

Carpaccio

XV

Girlandaio

Lorenzo Pinturiccio

XIV

Signorelli

Perujino

Boticelli

Giov.Bellini

Leonardo Titian

Giorgione

Michelangelo Rafael

PERUGIA

Tintoretto

XVI

Veronese

Course onFLORENCE KE, part 5, by Gavrilova T.

VENICE

33

«Living-room» mix ontology living_room.dom File E dit Insert V iew Table

Floor

Window

Help

Living-room

Room TV

Furniture

Electronics Decoration

Soft furniture

Musical Center

Picture Bookshelf

Chair Dinner table

Sofa

Wall

Armchair Book

Is-a is_on includes

Course on KE, part 5, by Gavrilova T.

34

Ontology Building Algorythm: 5 steps 1. 2. 3. 4. 5.

Glossary development Laddering Disintegration Categorization Refinement

Course on KE, part 5, by Gavrilova T.

35

Detailing • 1. Gather all the relevant information, select and verbalise all essential concepts (glossary). • 2. Reveal hierarchies among these concepts and represent them visually (trees). • 3. Detail the concepts via top-down strategy. • 4. Form meta-concepts via bottom-up strategy. • 5. Exclude repetitions, synonyms, excessiveness and contradictions. Discuss the structure with the expert and edit it. Divide resulting graph into levels and draw it. Course on KE, part 5, by Gavrilova T.

36

Gestalt (good form) Principles • Law of Pragnanz (M. Wertheimer)- organization of any structure in nature or cognition will be as good (regular, complete, balanced, or symmetrical) as the prevailing conditions allow (law of good shape).

• Law of Proximity – objects or stimuli that are viewed being close together will tend to be perceived as a unit.

• Law of Similarity – things that appear to have the same attributes are usually perceived as being a whole.

• Law of Inclusiveness (W.Kohler)- there is a tending

to perceive only the larger figure and not the smaller when it is embedded in a larger.

• Law of Parsimony – the simplest example is the best or

known as Ockham’s razor principle (14-th century): ``entities should

not be multiplied unnecessarily''.

Course on KE, part 5, by Gavrilova T.

37

Law of similarity

Course on KE, part 5, by Gavrilova T.

38

Towards beautiful ontology Harmony:

 Conceptual balance • • • •

Concepts of one level should be linked with the parent concept by one type of relationship such as is-a, or has part. The depth of the branches should be more or less the same (±2 nodes). The general outlay should be symmetrical. Cross-links should be avoided as much as possible.

 Clarity •

•

Minimizing the number of concepts according to Ockham’s razor principle (14-th century): ``entities should not be multiplied unnecessarily''. The maximal number of branches and the number of levels should follow Miller number (7±2) The type of relationship should be clear and obvious if the Course on KE, part 5, by Gavrilova T. 39 name of the relationship is missed.

About balance

Course on KE, part 5, by Gavrilova T.

40

Categorization mistakes A H

m

s

n

d

f

g

q

h

j

B w

k

e

l

m

r

t

p

Course on KE, part 5, by Gavrilova T.

y

u

i

o 41

Distance

Glossary Place

Conceptual str.

Personality Pace

level

Communicative methods

Theoretical aspects Psychological aspect

Knowledge Elicitation Gender Individual methods

Methods

Practice

Time

Methods Age

Active

Cognitive level Passive

Professional level Gnosiological aspect

Collective Textological

Interview

Contact level Common lang. Course on KE, part 5, by Gavrilova T.

Vocabulary 42

Linguistics

Teaching Ontology on Knowledge Elicitation

Course on KE, part 5, by Gavrilova T.

43

Test N 30 (My knowledge ontology)

Course on KE, part 5, by Gavrilova T.

44

Test N 31 (home task) Ontology design Create •partonomy (city, furniture, house) • geneology (economical theories, computer history) •taxonomy (books, meals, drinks, cars) •attributive structure (computer, tree, bicycle, room) •Mix ontology (university) Course on KE, part 5, by Gavrilova T.

45

Test N 33. Try to create ontology

An ancient Chinese classification system for animals (by Borhes): 1. Those that belong to the Emperor 2. Those that have four legs 3. Wild dogs 4. Those that are likely to break a jar 5. Those that resemble flies, at least from a distance 6. Those that behave in a crazy way 7. Embalmed animals 8. Tame animals 9. Uncountables 10. Those that are drawn with a very fine brush, made of camel hair 11. Mythical beasts 12. Piglets, nursed on milk Course on KE, part 5, by Gavrilova T. 46 13. Et cetera

5.3. Ontology formalization Classification of ontology languages General ontology languages FOL (First-order Logic) DL (Description Logics) Frame logic Web-centric ontology languages XML (eXtensible Markup Language, W3C) W3C: World Wide Web Consortium RDF (Resource Description Framework, W3C) DAML (DARPA Agent Markup Language, US) + OIL (Ontology Inference Layer, or Ontology Infrastructure Language, European OntoKnowledge) New standard is called OWL (Ontology Web Language, W3C) Course on KE, part 5, by Gavrilova T.

47

Description logic • Based on concepts (classes) and roles – Concepts (classes) are interpreted as sets of objects – Roles are interpreted as binary relations on objects • Decidable fragments of First Order Logic – Closely related to propositional modal logics • Key features of Description Logics are – Well defined semantics (they are logics) – Provision of inference services • Description Logic is characterised by set of constructors provided for building complex concepts and roles from simpler ones. Usually include at least: – Conjunction, disjunction, negation – Restricted (guarded) forms of quantification Course on KE, part 5, by Gavrilova T.

48

Frame • Frame logic – Frame-based systems – Network-based natural structure

• Major reasoning tasks: – – – –

Subsumption Anchoring (Classification) Inheritance: Finding definitions for specific concepts Axioms reasoning: • Finding relationships among concepts • Attachment procedure reasoning

• Representation and reasoning can be formalized by description logics Course on KE, part 5, by Gavrilova T.

49

Web-centric ontology languages (3rd generation) • First generation: handwritten HTML pages • Second generation: machine generated and/or active pages – Still intended for direct human processing – Reading, browsing, form-filling, etc. • Third generation: machine understandable/processable pages – Will enable intelligent services – Requires further levels of interoperability – Standards for semantics as well as syntax Course on KE, part 5, by Gavrilova T.

50

Web-centric ontology languages

Course on KE, part 5, by Gavrilova T.

51

XML • HTML-> XML-> SGML (Standard Generalized Markup Language ) • In XML, tags are not fixed - one can invent new tags to structure the information in a web page • DTD (Document type definition) defines the legal building blocks of an XML document; it defines the document structure with a list of legal elements • DTD -> XML schema Course on KE, part 5, by Gavrilova T.

52

XML • XML schema express shared vocabularies and allow machines to carry out rules made by people.; they provide a means for defining the structure, content and semantics of XML documents • XSL (eXtensible Stylesheet Language) for different presentation styles of XML documents • XML is considered to be the basis for all semantic web languages - the “machine code” of the new generation web

Course on KE, part 5, by Gavrilova T.

53

RDF • RDF was built on URl (uniform resource identifier)+ XML • Represent information about resources in the Web in terms of subject-predicate-object – Resource: subject, e.g. http://www.example.org/index.html – Property: predicate, e.g., creator – Property value: object, e.g., John Smith • RDF schema allows anyone to write their own name-space document (a ‘schema’); this defines properties and classes in some application domain • Weak in describing Web resources in sufficient detail Course on KE, part 5, by Gavrilova T.

54

OWL(DAML+OIL) •

Describes structure of the domain – RDF used to describe specific instances

•

Structure described in terms of classes (concepts) and properties (roles), just like Description Logic

•

Ontology consists of set of axioms – E.g., asserting class subsumption/equivalence

•

Classes can be names or expressions – Various constructors provided for building class expressions

•

Expressive power determined by – Kinds of axiom supported – Kinds of class (and property) constructor supported Course on KE, part 5, by Gavrilova T.

55

5.4. Ontologies on the Web • Ontolingua project • KA2 Initiative and Ontobroker

project • SHOE project

Course on KE, part 5, by Gavrilova T.

56

Ontolingua • Developed in the Knowledge System Laboratory - KSL http://www.ksl.stanford.edu.

• Program implementation of the Ontolingua system - Common Lisp.

• Main purpose – support of the user’s formal task’s specification on the basis of formal descriptions library of tasks’, models’ and concepts’ fragments and introduction of the fragments’ library.

• Knowledge Interchange Format (KIF) - language of Ontolingua system formal descriptions. Developed in KSL and proposed as a standard of intercomputer knowledge exchange. KIF language - the language of second order predicate calculus for representation of non procedural knowledge in which procedural knowledge is represented by rules of terms copying.

http://www-ksl-svc.stanford.edu Course on KE, part 5, by Gavrilova T.

57

Technology of Work with Ontolingua Ontolingua

User

Ontologies’ Library Ontology 1 Theory Classes Relations Functions Axioms …………………. Models’ fragments ... Concepts ...

Formalising of user’s task

Internet

Task’s theory: • variables, • classes, • relations, • functions, • axioms, • equations. Question.

Specialized servers of mathematical processing Logical inference Equations’ solving, Mathematica. Simulation modeling. Course on KE, part 5, by Gavrilova T.

58

User’s abilities in Ontolingua • Review of Ontolingua library sections represented in the form of HTML –documents; • Search for theories and definitions in the Ontolingua library; • Creation and editing of the user’s own ontology; • Syntactical validation of the created specifications and links to other ontologies; • Preparation of hypertext documentation on ontologies in the form of HTML documents; • In the presence of Common Lisp system - possibility of local operation with Ontolingua system. Course on KE, part 5, by Gavrilova T.

59

KA2 Initiative and Ontobroker project (Knowledge Annotation Initiative of the Knowledge Acquisition Community)

Main research trends: • Ontology engineering; • Web-pages annotation; • Inquiries for information on Web-pages and answers inference on the basis of ontology knowledge. http://ontobroker.aifb.uni-karlsruhe.de/ Course on KE, part 5, by Gavrilova T.

60

KA2 Initiative and Ontobroker project Sections of Common Ontology • organization ontology; • project ontology; • person ontology; • research-topic ontology; • publication ontology; • event ontology; • research-product ontology; • research-group ontology. Course on KE, part 5, by Gavrilova T.

61

KA2 Initiative and Ontobroker project

Course on KE, part 5, by Gavrilova T.

62

SHOE Project

(Simple HTML Ontology Extensions)

Main research trends: • Development of reusable ontologies for

concepts that are most frequent for Webresources; • Development of knowledge annotators.

http://www.cs.umd.edu/projects/plus/SHOE/ Course on KE, part 5, by Gavrilova T.

63

SHOE Architecture Annotations Knowledge annotator

Webpages

User Interface

Text Editor

Search SHOE

Represent

Subject domain interface

Knowledge base

Course on KE, part 5, by Gavrilova T.

64

Tools for Visual Ontology Design • • • • • • • • • • • • • • •

OntoEdit (http://www.ontoprise.de/com/start_download.htm) Apollo (http:// www.apollo.open.ac.uk) WebOnto (http://kmi.open.ac.uk/projects/webonto) SymOnotX (http://www.symontox.org/), WebODE (http://webode.dia.fi.upm.es/) OpenKnoMe (http://www.topthing.com OntoSaurus (http://www.isi.edu/isd/ontosaurus.html) OntoLingua Server (http://ontolingua.stanford.edu) PROTEGE (http://www.protege.stanford.edu) OilEd (Manchester, UK) for DAML+OIL http://www.ontology.jp Frodo (http://www.dfki.uni-kl.de/frodo/RDFSViz/) Hozo OE - Mizoguchi Tool ( http://www.ei.sanken.osaka-u.ac.jp/oe/oe3_download_en.html ) KAON (http://kaon.semanticweb.org/) Knowledge Builder (http:// spe.cgu.edu/faculty/facpages/annettesteinacker/index.html ) LinkKFactory (http://www.landc.be) CAKE/VITA http:// www.csa.ru/ailab

Course on KE, part 5, by Gavrilova T.

65

Relevant sites

• http://www.aiai.ed.ac.uk/project/enterprise/enterprise/ontology • http:// www.ontoweb.org • http://www.xml.com/2004/07/14/examples/

Course on KE, part 5, by Gavrilova T.

66

Ontology application • • • • • • • •

Knowledge engineering and management User modelling Multimedia search and retrieval Query processing Agent-based computing E-learning E-commerce Semantic Web Course on KE, part 5, by Gavrilova T.

67

Teaching Ontologies’ Taxonomy

O n t o lo g y of C o n c e p t io n s

H is t o r ic a l e v e n ts

P e r s o n a lit ie s

P a r a d ig m e s

Course on KE, part 5, by Gavrilova T.

M e th o d s T o o ls

68

Traditional Approach to Research • • • •

Problem definition Reviewing the literature Experimental and descriptive designs Data collection and analysis (for human studies) • Mathematical modelling • Programme implementation (for CS) • Writing up the findings Course on KE, part 5, by Gavrilova T.

69

Ontology-based Approach • • • •

Problem-ontology definition Ontology of reviewed approaches Experiment framework design Data structure design and ontology development • Mathematical modelling and main results ontology design • Programme implementation (for CS) • Structured writing up the findings Course on KE, part 5, by Gavrilova T.

70

From Knowledge Engineering to Ontology Engineering • first generation of KE tools - early 80-ies - the reconstruction of semantic space of human expertise and repertory grid-centred tools Expertise Transfer System (ETS) [Boose, 1986], AQUINAS [Boose, Shema, Bradshaw, 1989; Gaines, 1986] and others. • second generation KE tools - visual knowledge engineering provides ideas of CASE technology to AI [Aussenac-Gilles, Natta, 1993; Eisenstadt, Domingue,Motta, 1991]. • new generation of KE tools - special tools that help knowledge capture and structuring - last 5-7 years - KA tools that help to cut down the revise and review cycle time and to refine, structure and test human knowledge and expertise in the form of ontology [PROTEGE, WebOnto, OntoEdit, 2001-2003]. Course on KE, part 5, by Gavrilova T.

71

Public domain libraries that are accessible via Web • • • • • • •

DAML repository (http://www.daml.org/ontologies/). It is a repository with more that 160 ontologies that can be freely uploaded by the users. Ontologies are implemented in DAML+OIL language. Ontolingua Server repository (www-ksl-svc.stanford.edu:5915/). It stores more than 50 ontologies in Ontolingua. Universal repository (http://www.ist-universal.org/). The purpose of this repository is to enable collaboration among leading educators by providing exchange services for learning resources. SHOE repository (http://www.cs.umd.edu/projects/plus/SHOE/onts/). It provides descriptions and links to existing SHOE ontologies. WebODE (http://babage.dia.fi.upm.es/webode/) WebONTO (http://kmi.open.ac.uk/projects/webonto/) Ontosaurus (http://www.isi.edu/isd/ontosaurus.html)

Course on KE, part 5, by Gavrilova T.

72

Test N32 My project as a set of ontologies

Course on KE, part 5, by Gavrilova T.

73

Conclusion 1.

Оntologies are easy for the domains with good structure. 2. What about new and ill-structured disciplines as HCI, cognitive scinces, management,etc.? 3. Ontologies are very subjective. 4. Ontologies are good for knowledge sharing, reuse and teaching 5. They work as a mindtool! Course on KE, part 5, by Gavrilova T.

74